A Robust Watermarking Scheme Based on the Mean …downloads.hindawi.com/journals/scn/2018/1254081.pdfA Robust Watermarking Scheme Based on the Mean Modulation of DWT Coefficients ZiedKricha
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Research ArticleA Robust Watermarking Scheme Based on the Mean Modulationof DWT Coefficients
Zied Kricha 12 Anis Kricha34 and Anis Sakly5
1National School of Engineers of Sousse University of Sousse Erriadh City 4023 Tunisia2CSR Group with the Electronic and Microelectronic Lab National School of Engineers of MonastirUniversity of Monastir Ibn El Jazzar Avenue 5035 Tunisia
3LATIS Lab National School of Engineers of Sousse University of Sousse Erriadh City 4023 Tunisia4National School of Engineers of Monastir University of Monastir Ibn El Jazzar Avenue 5035 Tunisia5ESIR Research Unit National School of Engineers of Monastir University of Monastir Ibn El Jazzar Avenue 5035 Tunisia
Correspondence should be addressed to Zied Kricha krichaziedgmailcom
Received 4 May 2018 Revised 29 October 2018 Accepted 21 November 2018 Published 6 December 2018
Academic Editor Emanuele Maiorana
Copyright copy 2018 Zied Kricha et al This is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited
In an endeavor to deal with image copyright infringement robust watermarking approaches are commonly used However mostof the existing approaches either present a limited robustness or rely on highly computational algorithms thereby limiting theefficiency of these solutions In this paper a novel blind and robust watermarking method is presented First the vertical andhorizontal subbands coefficients resulting from the wavelet transformation are scrambled using a chaotic sequence and thengathered into individual blocks Next the mean value of each block is modulated according to watermark bit At the extractionstage based on the sign of the blocksrsquo mean a blind watermark extractor is suggested The imperceptibility security complexityand robustness of the proposed approach have been evaluated and compared with state of the art solutions Experimental resultsprove that the proposed approach successfully satisfies the watermarking requirement and outperforms existing methods againstboth geometric and signal processing attacks
1 Introduction
We all remember directing a banknote toward a source oflight to verify its authenticity The presence of a predefinedpattern number or portrait technically called watermarkdetermines whether the banknote is fake or not Watermark-ing is an old technique used to prevent counterfeiting ofseveral official paper documents Nowadays physical form ofpaper is not the only valuable thing worth securing whereasimages audio tracks and video sequences present a new formof money that is vulnerable to unauthorized distribution andmodification In an effort to address these issues watermark-ing concept was again adopted but this time with the prefixldquodigitalrdquo for referring to digital objects Digital watermarkingis the act of incorporating supplemental information intomultimedia material (still images audio or video) in such amanner that it does not alter the delivered material and yetstill readily exploited by the watermark detector even after
some undesirable operations An overview of the literaturereveals a broad range of application domains includingencryption [1] copyright protection [2] integrity verification[3] content restoration [4] and documentation [5]
The main focus of this paper is to design a robustwatermarking scheme for image copyright protection Thefirst watermarking schemes in this regard targeted the leastsignificant bit (LSB) of pixelsrsquo brightness [6] Despite theirsimplicity of realization LSB based schemes suffer from ahigh sensitivity to noise which prompted researchers tolook for new alternatives In fact researchers have targeteddifferent parts of the watermarking process the generation ofcarrier elements from the raw representation (eg pixels val-ues) [7 8] the choice of an embedding method summarizedby the embedding equation that describes how the water-marked elements are generated from the original coefficients[9 10] and finally the choice of a suitable watermark detector[11]
HindawiSecurity and Communication NetworksVolume 2018 Article ID 1254081 16 pageshttpsdoiorg10115520181254081
2 Security and Communication Networks
The two main used techniques for watermark embeddingare the spread spectrum (SS) [12] and the quantization(QIM) [10] based techniques These techniques have beensubject to much amelioration such as the spread transformdither modulation [13] and the progressive quantization [14]for QIM-based-techniques and the improved SS [9] themultiplicative SS (MSS) [15] and the improved MSS for SS-based-techniques [16]
The digital watermark may be inserted either in the rawrepresentation of the image where each watermark bit isinserted straight into pixelrsquos brightness or after a preprocess-ing phase involving the generation of a new domain orandthe selection of an appropriate component on which theembedding algorithm will be applied
The straightforward option though it is not the robusteris to insert the watermark in the spatial domain A pre-liminary work in this trend is presented in [17] where thewatermark is repeatedly inserted in a subtractive or additiveway depending on the pixelsrsquo characteristics To consolidatethe robustness of the DC coefficients with the rapidity ofdirect pixels manipulation Su et al [18] proposed to computethe DC without going through the DCT transformationAlthough this method is computationally lighter than theconventional insertion in transform domain it providesfewer choices and it is limited so far to the DC coefficients
In contrast to the spatial domain transform ones offermore flexibility For instance in the DCT domain thewatermark has been embedded into different positions suchas the DC [19] or the AC frequency [20] Similarly in theDiscrete Wavelet Transform (DWT) researchers opted fordifferent subbands and different wavelet families In [21]the watermark is embedded into the 1198671198713 subband of theHaar wavelet decomposition and into the 1198711198715 subband ofthe integer DWT in [22] Other employed transformationsinclude the gyrator domain [1] the contourlet [23] and theHessenberg [24] transforms
When most of schemes have been designed to embed thewatermark into the amplitude of coefficients either in thespatial or frequency domain some academics have chosento create or form a new carrier using various representationsof these coefficients Inspired by the works realized incommunication theory specifically the modulation of thephase instead of the amplitude Ourique et al [25] proposedto watermark the angle constituted by the host vector withrespect to a hyperspherical coordinate system
Recently many researchers have tended to embed thewatermark into a robust feature obtained from a set of coeffi-cients such as the coefficient difference computed betweenthe two largest wavelet coefficients belonging to the samegroup [7] or the difference between adjacent coefficients ofthe same DWT subbands [8] In [26 27] the watermarkis embedded into the imagersquos significant elements In [27]the nonsubsampled Shearlet transform of images significantregion is calculated and then the watermark is embedded inthe high-pass subband with the highest energy The humanvisual system can barely notice the modifications of highedge concentration zones To benefit from this weakness theauthors of [26] proposed to adjust the watermark strengthaccording to the blocks complexity The embedding stage
consists in the modification of two DCT coefficients fromeach block At the extraction phase the difference betweenthese coefficients specifies the value of the watermark bit In[28] Lilu et al took advantage of the LL subbandrsquos robustnessto provide blindness to the additive SS algorithm where theHH subband is replaced by the quantized and watermarkedLL subband
The bottom line of the literature review reveals that thewatermarking problem persists and the dilemma remainsunsolved Most of the existing schemes either manifest aweak robustness or make recourse to overcomputationalschemes through a cascade of transformations [14 29 30]metaheuristic algorithms [31] or both [32 33]
The main objective of this paper is to investigate andimplement an efficient watermarking scheme for grayscaleimages This objective is achieved by using a single trans-formation (DWT) to localize the appropriate frequencies ablock mean as a representation of the watermark carrier achaotic encryption for security and a mean modulation forembedding
To the best of our knowledge the closest work to oursis perhaps that of [34] in which the authors embedded thewatermark in the mean value of DWT coefficients Howeverthe difference between the two approaches is threefoldFirstly the two papers are dealing with different cover mediain [34] the host signal is an auditory signal while this paperaddresses robust image watermarking problem Secondly themean value in [34] is computed between adjacent coefficientsobtained from the same subband while in our work it is com-puted from gathered coefficients from different subbandsThirdly the embedding method is different where in [34]the mean value is moved to the nearest integer multipleof the used quantization step However in this work thewatermark is embedded by changing the sign of the meanvalue
The rest of this paper is organized as follows In Section 2we introduced the advantages of the used watermark carrierSection 3 describes how the watermark is inserted anddetected The performances of our scheme are evaluated anddiscussed in Section 4 Finally the conclusions are stated inSection 5
2 Watermark Carrier Selection
As discussed in the previous section the selection of awatermark carrier involves the choice of a transform domainalongside with an adequate representation
21 Transform Domain The wavelet transformation (WT)is a mathematical tool used to analyze a discrete or con-tinuous signal by extracting its frequency components andlocalize them in time Hence it allows access to importantinformation unattainable in the original representation TheWT is widely used in multidisciplinary fields such as astron-omy economy earthquake prediction signal processingand watermarking Although developed on the basis of theFourier transform the WT is more effective in terms ofcomputational complexity and frequency localization
Security and Communication Networks 3
64 128
256
512
1024
2048
4096
Image heightwidth (pixels)
0
10
20
30
40
50
60
70
80
90
Com
puta
tion
time (
s)
DWTDCTShearlet
2D Transformation
Figure 1 Computational time evolution with respect to the image dimension of some 2D transformations
In order to prove the complexity advantage of the DWTover other transformations the computational time of theDWT(1 level) DCT (8times 8 blocks) and the Shearlet transformare examined using different image sizes (Figure 1) Thesetransformations are employed in [7 28 35] [13 20 36]and [27] respectively The advantage of the DWT over othertransformations is clear for large image sizes and especiallyover the DCT which is the most time consuming among thethree evaluated transformations
Due to the above advantages we adopted the DWTtransformation in our schemeWhen applied to an image theDWT is achieved by filtering separately the rows and columnsof the image Each individual image transformation producesfour separable subbands one coarse subimage locating thelow frequencies (119871119871) and three locating the horizontal (119867119871)vertical (119871119867) and diagonal (119871119871) details (high frequencies)The WT may be repeated as appropriate by filtering againthe low frequency subband Each individual operation stageproduce a resolution level and each higher level yields to abetter frequency localization but a shrinkage in the subbandsize From the perspective of imperceptibility the119871119871 subbandis inappropriate for watermark insertion The approximationcoefficients represent the low frequencies that reflect signif-icant information about an image Thus the manipulationof these coefficients provokes quality degradation On theother hand the modification of the 119867119867 coefficients maygo unnoticed but the watermark may be easily removedby image processing tasks such as lossy compression Theremaining subbands (119867119871 and 119871119867) serve as a good candidatefor watermark embedding They are midfrequency subbandsthus they are less noticeable andmore robust than the 119871119871 and119867119867 subbands respectively
22 Carrier Selection Regardless of the insertion methodthere are always two ways to embed a watermark either to
insert one bit in a single coefficient or one bit in multiplecoefficients The latter solution leads to some importantadvantages of both robustness and imperceptibility [36]In fact a good watermark carrier has to guarantee twomain properties a high robustness combined with a lowperceptibility From a carrierrsquos point of view these prop-erties are manifested by a low distortion after attacks andan imperceptible change after embedding the watermarkThe blocksrsquo mean values (BMV) of 119867119871 and 119871119867 subbandssatisfy the above requirements it is computed from multiplecoefficients and it manifests a high robustness and imper-ceptibility To prove this claim we analyze the variation ofBMVs through the cumulative distribution function (CDF)where each block is constructed by gathering eight coef-ficients from unrepeated random positions The insertionof a watermark bit into a BMV can only be accomplishedthrough the modification of its block-coefficients Henceto emulate the insertion we propose to modify all thecoefficients of each block by a fixed value (negative orpositive)
First we investigate the quality of the resulting imageTheCDF in Figure 3(b) shows the large variations of the BMVsexpressed in terms of percentage the variation reaches upto 2 000 where 58 of the BMVs are modified between100 and 950 from their original values Despite this largevariation the modified image (Figure 2(b)) remains highlysimilar to the original one (Figure 2(a))
Concerning the robustness Figures 3(c) and 3(d) depictthe variation of BMVs always in terms of percentage afterJPEG compression and median filtering respectively Herewe can observe the limited variation which implies a highrobustness In fact in 72 of the cases the variation is lessthan 10 after compression and less than 20 in 73 of thetimes after filtering operations In both cases the variation isless than 30 in 83 of the BMVs
4 Security and Communication Networks
(a) (b)
Figure 2 Lena image (a) original and (b) after modification of all the1198671198711 and 1198711198671 coefficients
0010203040506070809
1
CDF
|BMV|
0 10 20 30 40 50 60 70 80
(a)
0 2000400 800 1200 1600BMV variation ()
0
01
02
03
04
05
06
07
08
09CD
F
(b)
BMV variation ()
0
01
02
03
04
05
06
07
08
09
CDF
0 10 20 30 40 50 60 70 80 90 100
(c)
0
01
02
03
04
05
06
07
08
09
BMV variation ()
CDF
0 10 20 30 40 50 60 70 80 90 100
(d)
Figure 3 Absolute value of Cumulative histogram of (a) original BMVs (b) difference between the original BMVs and the altered ones ((a)and (b)) (c) difference after JPEG(30) compression and (d) difference after median filtering (3 times 3)
It is worth noting that despite the robustness and imper-ceptibility of the BMVs the final results will also depend onthe insertion and detection method described in the nextsection
3 The Proposed Scheme
This section describes the insertion and extraction processesof a binary watermark 119872 = 1198871 1198872 119887119898 in a grayscale
Security and Communication Networks 5
2timesℎ
2 times w
LHl
HLl
c1 c2 c3 cn
c1 c2 c3 cn
x0
b1 b2 b3 bm
Original Image
Host coefficientsShuffling Chaotic sequence
generation
Block partitioning
Block meanmodulation
Watermark bits
Reshuffling
Watermarked subbands
Watermarked coefficients
BL1 BL2BLlfloorntrfloor
BL1 BL
2 BL
m
BLm+1BLlfloorntrfloor
LHl
HLl
c1c2
c3cn
Figure 4 Block diagram of the watermark embedding scheme
image I with 2 times ℎ height and 2 times 119908 width At this point theembedding parameters are not specified but kept generalizedto point out the flexibility of the scheme At first the originalimage goes through 119897 levels of DWT in order to manifest thedifferent frequencies of the image For the reasons presentedin Section 2 the 119871119867119897 and the 119867119871 119897 subbands are selected forembedding Let 119871119867119897 = 119871119867(119894 119895) and 119867119871 119897 = 119867119871(119894 119895) bethe full set of carrier coefficients where 119894 isin [1 1199082119897minus1] and119895 isin [1 ℎ2119897minus1] The coefficients of the two subbands are thenconcatenated into a single vector 119862 = 119888119896 where 119896 isin [1 119899]and 119899 = ℎ times 11990822119897minus3 Next the carrier vector 119862 is shuffledusing a chaotic sequence generated using the logistic mapand eachwatermark bit is embedded bymodulating themeanvalue of a set of coefficients The complete block diagramof the embedding process is given in Figure 4 The threemain stages involved in the watermarking process (chaoticsequence generation coefficients shuffling and mean blockmodulation) are detailed in the rest of this section
31 Chaotic Sequence Generation Chaotic maps are dynamicsystems defined by a mathematical function ((119909119899+1 = 119891(119909119899))that model a chaotic behavior In practice a variety of chaoticsystems may be utilized (eg Bakerrsquos map circle map duffingmap logistic map etc) The implementation of the latter isone of the simplest thanks to its polynomial mapping degreethough itrsquos high dynamic complexity These features allowedthe logistic map to attract a lot of attention in the informationsecurity disciplines such as cryptography and watermarking
The logistic map may be described by the followingrecurrence relation
119909119899+1 = 119891 (119909119899) = 120582 (119909119899 minus 1199092119899) (1)
The chaotic sequence 119909119899 generated by the function 119891 after119899 iterations is determined by its input parameters 120582 and 1199090where 1199090 is the initial value of the sequence and 120582 is thecontrol parameters In practice 1199090 is choosing between [0 1]and120582 lies between 3569955672 and 4These ranges guaranteea chaotic sequence with a high randomness in the outputwhere the slightest difference in the initial parameters (1199090 and120582) will lead to uncorrelated sequences over time
32 Shuffling In the shuffling phase the elements of the vec-tor 119862 are assigned to new positions according the generatedchaotic sequence The objective of this shuffling is double-fold Firstly it aims to improve the security of the schemethe 119905 coefficients forming each block (119861119871) are randomlyselected therebymaking it difficult for a third party to extractthe watermark or selectively attack the host coefficients todestroy the watermark without the permutation sequenceSecondly the shuffling dispels the correlation between adja-cent coefficients In fact adjacent coefficients tend to beof the same sign leading to MVBs distant from zero Incontrast the MVBs of scattered coefficients are gatheredaround the zero value which will serve us to minimizethe embedding strength and hence the perceptibility of thewatermark
The logistic map defined by (1) is employed to producea pseudo-random key containing the new positions for eachcoefficient
Let 119909119899 = (1199091 1199092 1199093 119909119899) be the chaotic sequencegenerated by means of the logistic map using the secretkey SK = 1199090 120582 where 119899 is the number of iterationsthat is equal to the total number of coefficients in 119871119867119897 and119867119871 119897 The 119909119899 coefficients are then sorted in a ascending
6 Security and Communication Networks
LL
HH
12
Chaotic sequence
4 2 7 1 6 5 3 8
12 09 -03 01 -24 -11 -07 -18
01 09 -07 12 -11 -24 -03 -18
022 025 028 067 074 081 090 0914 2 7 1 6 5 3 8
067 025 090 022 081 074 028 0911 2 3 4 5 6 7 8
Permutation sequence
Sorting09
-03 01
-24 -11
-07 -18C
C
Figure 5 Shuffling of DWT diagonal subbands using a chaotic sequence of eight real numbers
order (1199091199041 1199091199042 1199091199043 119909119904119899) and so the permutation sequence(1199041 1199042 1199043 119904119899) is obtained
The example shown in Figure 5 illustrates a concreteexample for the shuffling operation In this example 119899 is equalto eight the vector 119862 is the concatenation of the diagonalDWT coefficients and 119862 is the shuffled version of 11986233 Block Mean Modulation The first step in the blockmodulation phase is the partitioning of the confused hostcoefficients 119888119894 into independent blocks This operation isrealized by gathering each 119905 adjacent coefficients into a singleblock 119861119871 119894 as given by (2) where 119894 is the index of the blockand is the concatenation symbol The total number ofgenerated blocks 119861119871 should be larger or equal than the binarywatermark message length In other words lfloor119899119905rfloor ge 119898 suchas 119905 isin Nlowast
To embed a robust watermark we consider the quantizationtechnique which modulates the carrier hosts to 119875 nonover-lapping sets where 119875 represents the possible embeddedwords In this work the watermark is a binary messagehence the carrier will be modulated to two different sets(119901 = 2) each associated to a watermark bit The robustnessof the quantization-based insertion method depends on thedistance between the two sets On the other hand wideningthe gap will lead tomediocre results in terms of visual qualityTo work around this issue Chen et al [10] proposed toincrease the element of each set and it is up to the embeddingalgorithm to choose the appropriate quantizer among theavailable ones However as illustrated in Figure 3(a) mostof the BMV are concentrated around the zero where in 80of the BMVs are in the range [minus20 20] Thus we propose tomodulate the blocksrsquo mean values by a positive or negativenumber according to the embedded bit Let 119861119871 119894 be the meanvalue of the block 119894 defined by
119861119871 119894 = 1119905119894times119905sum
119895=119905(119894minus1)+1
119888119895 (3)
The modulation of 119861119871 119894 goes by firstly nullifying the meanvalue by spreading its value equally to all blockrsquos coefficientsthen adding or subtracting a threshold 119879 according to theembedded bit 119887 The equation of modulation is given by (4)where 119888119896 is a coefficient belonging to the block 119894 (119861119871 119894)
1198881015840119896 = 119888119896 minus 119861119871 119894119905 minus 119879 if 119887 = 0119888119896 minus 119861119871 119894119905 + 119879 otherwise (4)
Once all the watermark bits are inserted the watermarkedcoefficients are rearranged into their initial positions usingthe same key SK Finally the inverse DWT is performed toobtain the watermarked image
34 Watermark Extraction In the watermark extractionphase the watermarked image is firstly transformed in thefrequency domain as in the embedding and the same key SKis used to constitute different blocks The decision upon theembedded bit is judged according to the blockrsquos mean valueof the attacked watermarked block 119861119871119894
119887 = 1 if 119861119871119894 gt 00 otherwise (5)
4 Experimental Results
This section is dedicated to evaluate the performance ofthe suggested watermarking framework A variety of 8-bitdepth (119889 = 8) grayscale standard benchmark images witha resolution of 512 times 512 pixels (119908 = ℎ = 256) have beensubjects of assessment The used image database includesLena Baboon Peppers and Tank These images present avariety of image properties (eg mixture of dark and lightarea alongside with highly textures regions (hair) in the Lenaimage) which refutes the idea of tying the performance of thescheme to some image properties
The performance of the suggested method is experimen-tally examined from different aspects Firstly the relation
Security and Communication Networks 7
38 41 45PSNR (dB)
2
4
6
8
10
12
14
16
18
Thre
shol
d va
lue (
T)
123
DWT level
Figure 6 Relation of threshold value (T) and the PSNR at different DWT levels
between the capacity and the imperceptibility is evaluatedusing different insertion parameters Next the evolution ofcomplexity according to the host image size is monitoredAfterwards the robustness is validated in two ways initiallythe robustness against different levels of attacks is examinedand then a comparative study with similar works is presentedFinally the security of the scheme is verified
41 Imperceptibility The imperceptibility feature reflects thefidelity of the watermarked image toward the original oneSeveral metrics can be used to quantify the noticeable differ-ence between two images involving subjective and objectivemetrics In this work we employed the Peek Signal to NoiseRatio (PSNR) and the Structural Similarity Index Measure(SSIM) given by (6) and (7) respectively
PSNR (I Ilowast) = 10 log10( (2119889 minus 1)2 times 119908 times ℎsum119908119894=1sumℎ119894=1 (I119894119895 minus Ilowast119894119895)2) (6)
where 120579I and 120583I are respectively the mean intensity and thestandard deviation of the original image I and 120590119868119868lowast is thesample cross-correlation between I and Ilowast The constants 1198781and 1198782 are defined as 1198781 = (1198961(2119889minus1))2 and 1198782 = (1198962(2119889minus1))2where 1198961 1198962 ≪ 142 Capacity The maximum capacity of the proposedscheme depends on three variables the image dimension (119908
and ℎ) the wavelet level (119897) and the block length (119905) Hencewe can express the maximum capacity using
Max capacity = lfloor 119908 times ℎ119905 times 22119897minus3rfloor (8)
For a better understanding of the relation between the capac-ity and imperceptibility we embedded a variable message size(119898 isin 256 512 1024) at different levels of decomposition(119897 isin 1 2 3) into three different images namely Lena Coupleand Baboon For each watermarking operation the blocksrsquosize 119905 is maximized so that all coefficients are involved inthe watermarking process (eg 119897 = 2 and 119905 = 32 for119898 = 1024) The threshold 119879 is manually chosen to reach aPSNR value of 38 41 and 45 dBThe parameters of insertion(119879 119905 119897 119898) the objective measurements (PSNR and SSIM)and the watermarked images are all listed in Table 1
The first observation that we draw from Table 1 is thatfor a given PSNR the threshold 119879 is converging towardthe same value regardless the host image or the watermarksize This amounts to saying that we can anticipate the119879 value according to the level of transformation and thetargeted PSNR Figure 6 illustrates the relation between theseparameters where119879 is obtained by averaging all the values fora specific level and PSNR Secondly it is shown from Table 1that at 45 dB the distortion due to the watermark insertionis invisible whatever the size of the message and the levelof decomposition However for watermarked images with aPSNR of 38 and 41 dB the imperceptibility is conditionedupon the level 119897 the higher 119897 is the lower the quality is goingto be
43 Complexity The computational cost of the watermarkembedding and extraction is an important aspect that may
8 Security and Communication Networks
Table 1 Quality of watermarked images using different insertion parameters (119897 119879119898)Message size Transformation
level Watermarked Lena images Watermarked Coupleimages
Figure 7 Computational time of (a) overall insertion and detection and (b) details of the embedding stage
restrict or expand the use of a given watermarking schemeIn the literature most of schemes are evaluated in a standardimage with a size of 512 times 512 pixels and there is noperspective of the computational cost trend when the imagegoes bigger which is more realistic Hence we propose tovary the size of the image beginning with 64 times 64 pixelsand ending with 4096 times 4096 pixels by doubling the heightand width after each insertion For the sake of simplicity thewatermark was embedded after one level of DWT (119897 = 1)with a fixed block size (119905 = 4) and a maximum watermarksize as defined by (8)
Figure 7(a) depicts the execution time evolution ofthe overall embedding and detection alongside with thevariation of the watermark size (red) as a function of theimage size The details of the embedding phase are shownin Figure 7(b) The most important remark to be made fromFigure 7 is that the computational time evolves broadly inline with the watermark size (red curve) to find out thecomplexity of our scheme in comparison with other state-ofthe-art methods
44 Robustness In the following we investigate the robust-ness of the suggested watermarking approach under severaltypes of attacks The robustness is quantified using the BitError Rate (BER) and the Normalized Correlation (NC)metrics
441 Evaluated Attacks Since the proposed scheme isdesigned to be robust the embedded message should lastor at list a significant part of it after image processingoperations Fromwatermark perspective all image processingoperations whatever performed on goodwill or not is consid-ered as attacks The involved attacks in the evaluation tests areclassified into three groups
(i) Compression attacks usually images are not storedin a raw format but after a compression operationthat aims to reduce the required storage space JPEGalgorithm is one of the most used compression
algorithms However JPEG is a lossy algorithm thatcauses some quality degradation inversely commen-surate with the compression quality factor Hence wetested the robustness of our algorithm under differentJPEG quality factors ranging from 10 to 90
(ii) Image processing attacks used to imitate a noisecontaminating the image during transmission such assalt and peppers or to improve the imagersquos qualityby eliminating a noise or enhancing some propertiesThe used attacks are listed as follows scaling medianfiltering sharpening Gaussian filtering average fil-tering Gaussian noise salt and pepper noise contam-ination histogram equalization and additive whiteGaussian noise (AWGN)
(iii) Geometric attacks used to correct geometric distor-tions such camera orientation (rotation) or imagesrsquoalignment (cropping) The used attacks are listed asfollows image cropping and resizing and rotation
Since the DWT coefficients are not geometrically invari-ant a manual synchronization of the imagersquos dimension andorientation was realized prior to the extraction
442 Performance Evaluation Here we test the robustnessof our scheme on three different images Boat Peppersand Barbara Each time we varied the embedded messagebetween 128 bits and 2048 bits The embedding threshold 119879is manually tuned to provide a watermarked image with aPSNR value of 40 dB and 45 dB The robustness is expressedin terms of BER () after mean filter (Figure 8) JPEGcompression (Figure 9) median filter (Figure 10) and gammacorrection Poisson noise histogram equalization and imageunsharpening (Figure 11) Figures 8ndash11 show that the robust-ness performance slightly differs between images For a smallmessage length (between 128 and 256 bits) and a PSNR valueof 40 dB the BER does not exceed 20 for most of attacksHowever the larger the message size is or the stronger theattack is the higher the BER will be
10 Security and Communication Networks
128 256 512 1024 2048Watermark size
0
10
20
30
40
50BE
R (
)
Mean filter (3x3)Mean filter (5x5)Mean filter (7x7)Mean filter (9x9)BarbaraPeppersBoat
(a)
Mean filter (3x3)Mean filter (5x5)Mean filter (7x7)Mean filter (9x9)BarbaraPeppersBoat
128 256 512 1024 2048Watermark size
0
10
20
30
40
50
BER
()
(b)
Figure 8 BER() after mean filter attacks using different message lengths at different PSNR values (a) 40 dB and (b) 45 dB
Figure 9 BER() after JPEG compression using different message lengths at different PSNR values (a) 40 dB and (b) 45 dB
443 Comparative Study To prove the superiority and therobustness advantages of our scheme a comparative studywith state of the art works is indispensable In fact wecompared the robustness of our scheme with six robustwatermarking schemes [26 27 35 37ndash39] In order to carryout a fair comparison the same insertion parameters (PSNRwatermark size) and attacks are used in this comparisonThe PSNRs value of different images and their correspondingvalues in the comparative schemes are illustrated in Figure 12It is worth mentioning that the insertion parameters (119897 119905 119879)were empirically chosen to produce the same PSNR alongsidewith an acceptable visual quality
The robustness results are presented in Tables 2ndash5 interms of either BER or NC depending on how they appearin the reference papers and the superior results are markedin bold number
Based on the simulation results given in Tables 2ndash5 it canbe found that our approach outperforms similar schemes on7194 of the time Equal results are obtained on 2343 ofthe time and our scheme underperforms on 462 of thecases
45 Security The robustness of a watermarking scheme isnot sufficient to judge a scheme as reliable As discussedin the first section robust watermarking is used to secureimages Hence the security is necessary First the watermarkshould be immune against illegal extraction even when theembedding procedure is exposed Second the watermarkdetector should not report the existence of the watermarkwhen none is embedded To evaluate these two risks weextract the watermark from a watermarked image using
Security and Communication Networks 11
Table 2 BER comparison with [27] and [37] using a message length of 128 bits
different keys and extract a watermark from a nonwater-marked image respectively In the two cases the maximumcorrelation obtained has not exceeded the value 059 whichpoints out a negative presence
5 Conclusion
The mean value of DWTrsquos vertical and horizontal subbandscoefficients is highly robust and imperceptible In fact these
two advantages were exploited in the design of a robustand blind watermarking approach The embedding methodconsisted in modulating the BMV in accordance with thequantization technique The use of a single transformationdomain alongside with a simple embedding technique makesfrom the proposed algorithmcomputationally undemandingAt the extraction phase the watermark bits were identifiedby the BMV position with respect to the zero value Unlikesome existing schemes the extraction rules did not require to
Security and Communication Networks 15
be adjusted to the applied attacks to fulfill a high robustnessIn fact the inserted watermark was capable to withstand withno error after JPEG compression up to a quality factor of 30Gaussian filter and other attacks
Data Availability
The data used to support the findings of this study areincluded within the article
Conflicts of Interest
The authors declare that they have no conflicts of interest
References
[1] S Liansheng Z Bei W Zhanmin and T Ailing ldquoAn opticalcolor image watermarking scheme by using compressive sens-ingwith human visual characteristics in gyrator domainrdquoOpticsand Lasers in Engineering vol 92 pp 85ndash93 2017
[2] D Shukla and M Sharma ldquoA Novel Scene-Based Video Water-marking Scheme for Copyright Protectionrdquo Journal of IntelligentSystems vol 27 no 1 pp 47ndash66 2018
[3] A Tiwari and M Sharma ldquoAn efficient vector quantizationbased watermarking method for image integrity authentica-tionrdquo Advances in Intelligent Systems and Computing vol 519pp 215ndash225 2018
[4] S Bravo-Solorio F Calderon C-T Li and A K Nandi ldquoFastfragile watermark embedding and iterative mechanism withhigh self-restoration performancerdquo Digital Signal Processingvol 73 pp 83ndash92 2018
[5] A Ghardallou A Kricha A Sakly and A Mtibaa ldquoAdaptiveblock sized reversible watermarking scheme based on integertransformrdquo in Proceedings of the 17th International Conferenceon Sciences and Techniques of Automatic Control and ComputerEngineering STA 2016 pp 347ndash351 Tunisia December 2016
[6] R G van Schyndel A Z Tirkel and C F Osborne ldquoA digitalwatermarkrdquo in Proceedings of the IEEE International ConferenceImage Processing (ICIP rsquo94) vol 2 pp 86ndash90 Austin Tex USANovember 1994
[7] W-H Lin S-J Horng T-W Kao P Fan C-L Lee and Y PanldquoAn efficient watermarking method based on significant differ-ence of wavelet coefficient quantizationrdquo IEEE Transactions onMultimedia vol 10 no 5 pp 746ndash757 2008
[8] S Bhattacharyya and G Sanyal ldquoA Robust Image Steganogra-phy using DWT Difference Modulation (DWTDM)rdquo Interna-tional Journal of Computer Network and Information Securityvol 4 no 7 pp 27ndash40 2012
[9] H S Malvar and D A Florencio ldquoImproved spread spectruma new modulation technique for robust watermarkingrdquo IEEETransactions on Signal Processing vol 51 no 4 pp 898ndash9052003
[10] B Chen and G W Wornell ldquoProvably robust digital water-markingrdquo in Proceedings of the 1999 Multimedia Systems andApplications II pp 43ndash54 September 1999
[11] Z Kricha A Kricha and A Sakly ldquoAccommodative extractorfor QIM-based watermarking schemesrdquo IET Image Processing2018
[12] I J Cox J Kilian T Leighton and T Shamoon ldquoSecure imper-ceptable yet perceptually salient spread spectrum watermark
formultimediardquo in Proceedings of the 1996 Southcon Conferencepp 192ndash197 June 1996
[13] X Li J Liu J Sun Z Sun H Hu and L Zhang ldquoImprovedspread transform dither modulation-based watermark algo-rithm based on step projectionrdquo in Proceedings of the 2010 IEEEInternational Symposium on Broadband Multimedia Systemsand Broadcasting (BMSB) pp 1ndash5 Shanghai China March2010
[14] H-T Hu and L-Y Hsu ldquoExploring DWT-SVD-DCT featureparameters for robust multiple watermarking against JPEG andJPEG2000 compressionrdquo Computers and Electrical Engineeringvol 41 pp 52ndash63 2015
[15] J Cannons and P Moulin ldquoDesign and statistical analysis of ahash-aided image watermarking systemrdquo IEEE Transactions onImage Processing vol 13 no 10 pp 1393ndash1408 2004
[16] A Valizadeh and Z J Wang ldquoAn improved multiplicativespread spectrum embedding scheme for data hidingrdquo IEEETransactions on Information Forensics and Security vol 7 no4 pp 1127ndash1143 2012
[17] M Kutter D Frederic and F B JordanDigital signature of colorimages using amplitude modulation 1997
[18] Q Su and B Chen ldquoRobust color image watermarking tech-nique in the spatial domainrdquo Soft Computing vol 22 no 1 pp91ndash106 2018
[19] Z Yanling Z Xiaoshi L Na et al ldquoA Digital ImageWatermarkAlgorithm Based onDCCoefficientsQuantizationrdquo in Proceed-ings of the 2006 6th World Congress on Intelligent Control andAutomation pp 9734ndash9738 Dalian China 2006
[20] RMZhaoH LianHWPang andBNHu ldquoAWatermarkingAlgorithm by Modifying AC Coefficies in DCT Domainrdquo inProceedings of the 2008 International Symposium on InformationScience and Engineering (ISISE) pp 159ndash162 Shanghai ChinaDecember 2008
[21] B L Gunjal and R RManthalkar ldquoDiscreteWavelet Transformbased strongly robust watermarking scheme for informationhiding in digital imagesrdquo in Proceedings of the 3rd InternationalConference on Emerging Trends in Engineering and TechnologyICETET 2010 pp 124ndash129 India November 2010
[22] H Shi and F Lv ldquoA Blind Digital Watermark Techniquefor Color Image Based on Integer Wavelet Transformrdquo inProceedings of the 2010 International Conference on BiomedicalEngineering and Computer Science (ICBECS) pp 1ndash4 WuhanChina April 2010
[23] S Ranjbar F Zargari and M Ghanbari ldquoA highly robust two-stage Contourlet-based digital image watermarking methodrdquoSignal Processing Image Communication vol 28 no 10 pp1526ndash1536 2013
[24] Q Su and B Chen ldquoA novel blind color image watermarkingusing upper Hessenbergmatrixrdquo AEU - International Journal ofElectronics and Communications vol 78 pp 64ndash71 2017
[25] F Ourique V Licks R Jordan and F Perez-Gonzalez ldquoAngleQIM a novel watermark embedding scheme robust againstamplitude scaling distortionsrdquo in Proceedings of the IEEE Inter-national Conference on Acoustics Speech and Signal Processing(ICASSP rsquo05) pp 797ndash800 March 2005
[26] H R Fazlali S Samavi N Karimi and S Shirani ldquoAdaptiveblind image watermarking using edge pixel concentrationrdquoMultimedia Tools and Applications vol 76 no 2 pp 3105ndash31202017
[27] X-Y Wang Y-N Liu H Xu A-L Wang and H-Y YangldquoBlind optimum detector for robust image watermarking in
16 Security and Communication Networks
nonsubsampled shearlet Domainrdquo Information Sciences vol372 pp 634ndash654 2016
[28] X Liu C Lin and S Yuan ldquoBlind Dual Watermarking forColor Imagesrsquo Authentication and Copyright Protectionrdquo IEEETransactions on Circuits and Systems for Video Technology vol28 no 5 pp 1047ndash1055 2018
[29] Poonam and S M Arora ldquoA DWT-SVD based Robust DigitalWatermarking for Digital Imagesrdquo Procedia Computer Sciencevol 132 pp 1441ndash1448 2018
[30] S Fazli and M Moeini ldquoA robust image watermarking methodbased on DWT DCT and SVD using a new technique forcorrection of main geometric attacksrdquo Optik - InternationalJournal for Light and ElectronOptics vol 127 no 2 pp 964ndash9722016
[31] K Kuppusamy and K Thamodaran ldquoOptimized Image Water-marking Scheme Based On PSOrdquo Procedia Engineering vol 38pp 493ndash503 2012
[32] V S Rao R S Shekhawat and V K Srivastava ldquoA dwt-dct-svd based digital image watermarking scheme using particleswarm optimizationrdquo in Proceedings of the 2012 IEEE StudentsrsquoConference on Electrical Electronics and Computer Science pp1ndash4 2012
[33] T T Takore P R Kumar and G L Devi ldquoA modified blindimage watermarking scheme based on DWT DCT and SVDdomain using GA to optimize robustnessrdquo in Proceedings ofthe 2016 International Conference on Electrical Electronics andOptimization Techniques ICEEOT 2016 pp 2725ndash2729 IndiaMarch 2016
[34] H Hu S Lin and L Hsu ldquoEffective blind speech watermarkingvia adaptive mean modulation and package synchronization inDWT domainrdquo EURASIP Journal on Audio Speech and MusicProcessing vol 2017 no 1 2017
[35] W-H Lin Y-R Wang S-J Horng T-W Kao and Y Pan ldquoAblind watermarking method using maximum wavelet coeffi-cient quantizationrdquo Expert Systems with Applications vol 36no 9 pp 11509ndash11516 2009
[36] X Li J Liu J Sun X Yang andW Liu ldquoStep-projection-basedspread transform dither modulationrdquo IET Information Securityvol 5 no 3 pp 170ndash180 2011
[37] M Hamghalam S Mirzakuchaki andM A Akhaee ldquoGeomet-ric modelling of the wavelet coefficients for image watermark-ing using optimum detectorrdquo IET Image Processing vol 8 no 3pp 162ndash172 2014
[38] V S Verma R K Jha and A Ojha ldquoDigital watermarkextraction using support vector machine with principal com-ponent analysis based feature reductionrdquo Journal of VisualCommunication and Image Representation vol 31 pp 75ndash852015
[39] C Li Z Zhang Y Wang B Ma and D Huang ldquoDithermodulation of significant amplitude difference for waveletbased robust watermarkingrdquo Neurocomputing vol 166 pp404ndash415 2015
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Navigation and Observation
International Journal of
Hindawi
wwwhindawicom Volume 2018
Advances in
Multimedia
Submit your manuscripts atwwwhindawicom
2 Security and Communication Networks
The two main used techniques for watermark embeddingare the spread spectrum (SS) [12] and the quantization(QIM) [10] based techniques These techniques have beensubject to much amelioration such as the spread transformdither modulation [13] and the progressive quantization [14]for QIM-based-techniques and the improved SS [9] themultiplicative SS (MSS) [15] and the improved MSS for SS-based-techniques [16]
The digital watermark may be inserted either in the rawrepresentation of the image where each watermark bit isinserted straight into pixelrsquos brightness or after a preprocess-ing phase involving the generation of a new domain orandthe selection of an appropriate component on which theembedding algorithm will be applied
The straightforward option though it is not the robusteris to insert the watermark in the spatial domain A pre-liminary work in this trend is presented in [17] where thewatermark is repeatedly inserted in a subtractive or additiveway depending on the pixelsrsquo characteristics To consolidatethe robustness of the DC coefficients with the rapidity ofdirect pixels manipulation Su et al [18] proposed to computethe DC without going through the DCT transformationAlthough this method is computationally lighter than theconventional insertion in transform domain it providesfewer choices and it is limited so far to the DC coefficients
In contrast to the spatial domain transform ones offermore flexibility For instance in the DCT domain thewatermark has been embedded into different positions suchas the DC [19] or the AC frequency [20] Similarly in theDiscrete Wavelet Transform (DWT) researchers opted fordifferent subbands and different wavelet families In [21]the watermark is embedded into the 1198671198713 subband of theHaar wavelet decomposition and into the 1198711198715 subband ofthe integer DWT in [22] Other employed transformationsinclude the gyrator domain [1] the contourlet [23] and theHessenberg [24] transforms
When most of schemes have been designed to embed thewatermark into the amplitude of coefficients either in thespatial or frequency domain some academics have chosento create or form a new carrier using various representationsof these coefficients Inspired by the works realized incommunication theory specifically the modulation of thephase instead of the amplitude Ourique et al [25] proposedto watermark the angle constituted by the host vector withrespect to a hyperspherical coordinate system
Recently many researchers have tended to embed thewatermark into a robust feature obtained from a set of coeffi-cients such as the coefficient difference computed betweenthe two largest wavelet coefficients belonging to the samegroup [7] or the difference between adjacent coefficients ofthe same DWT subbands [8] In [26 27] the watermarkis embedded into the imagersquos significant elements In [27]the nonsubsampled Shearlet transform of images significantregion is calculated and then the watermark is embedded inthe high-pass subband with the highest energy The humanvisual system can barely notice the modifications of highedge concentration zones To benefit from this weakness theauthors of [26] proposed to adjust the watermark strengthaccording to the blocks complexity The embedding stage
consists in the modification of two DCT coefficients fromeach block At the extraction phase the difference betweenthese coefficients specifies the value of the watermark bit In[28] Lilu et al took advantage of the LL subbandrsquos robustnessto provide blindness to the additive SS algorithm where theHH subband is replaced by the quantized and watermarkedLL subband
The bottom line of the literature review reveals that thewatermarking problem persists and the dilemma remainsunsolved Most of the existing schemes either manifest aweak robustness or make recourse to overcomputationalschemes through a cascade of transformations [14 29 30]metaheuristic algorithms [31] or both [32 33]
The main objective of this paper is to investigate andimplement an efficient watermarking scheme for grayscaleimages This objective is achieved by using a single trans-formation (DWT) to localize the appropriate frequencies ablock mean as a representation of the watermark carrier achaotic encryption for security and a mean modulation forembedding
To the best of our knowledge the closest work to oursis perhaps that of [34] in which the authors embedded thewatermark in the mean value of DWT coefficients Howeverthe difference between the two approaches is threefoldFirstly the two papers are dealing with different cover mediain [34] the host signal is an auditory signal while this paperaddresses robust image watermarking problem Secondly themean value in [34] is computed between adjacent coefficientsobtained from the same subband while in our work it is com-puted from gathered coefficients from different subbandsThirdly the embedding method is different where in [34]the mean value is moved to the nearest integer multipleof the used quantization step However in this work thewatermark is embedded by changing the sign of the meanvalue
The rest of this paper is organized as follows In Section 2we introduced the advantages of the used watermark carrierSection 3 describes how the watermark is inserted anddetected The performances of our scheme are evaluated anddiscussed in Section 4 Finally the conclusions are stated inSection 5
2 Watermark Carrier Selection
As discussed in the previous section the selection of awatermark carrier involves the choice of a transform domainalongside with an adequate representation
21 Transform Domain The wavelet transformation (WT)is a mathematical tool used to analyze a discrete or con-tinuous signal by extracting its frequency components andlocalize them in time Hence it allows access to importantinformation unattainable in the original representation TheWT is widely used in multidisciplinary fields such as astron-omy economy earthquake prediction signal processingand watermarking Although developed on the basis of theFourier transform the WT is more effective in terms ofcomputational complexity and frequency localization
Security and Communication Networks 3
64 128
256
512
1024
2048
4096
Image heightwidth (pixels)
0
10
20
30
40
50
60
70
80
90
Com
puta
tion
time (
s)
DWTDCTShearlet
2D Transformation
Figure 1 Computational time evolution with respect to the image dimension of some 2D transformations
In order to prove the complexity advantage of the DWTover other transformations the computational time of theDWT(1 level) DCT (8times 8 blocks) and the Shearlet transformare examined using different image sizes (Figure 1) Thesetransformations are employed in [7 28 35] [13 20 36]and [27] respectively The advantage of the DWT over othertransformations is clear for large image sizes and especiallyover the DCT which is the most time consuming among thethree evaluated transformations
Due to the above advantages we adopted the DWTtransformation in our schemeWhen applied to an image theDWT is achieved by filtering separately the rows and columnsof the image Each individual image transformation producesfour separable subbands one coarse subimage locating thelow frequencies (119871119871) and three locating the horizontal (119867119871)vertical (119871119867) and diagonal (119871119871) details (high frequencies)The WT may be repeated as appropriate by filtering againthe low frequency subband Each individual operation stageproduce a resolution level and each higher level yields to abetter frequency localization but a shrinkage in the subbandsize From the perspective of imperceptibility the119871119871 subbandis inappropriate for watermark insertion The approximationcoefficients represent the low frequencies that reflect signif-icant information about an image Thus the manipulationof these coefficients provokes quality degradation On theother hand the modification of the 119867119867 coefficients maygo unnoticed but the watermark may be easily removedby image processing tasks such as lossy compression Theremaining subbands (119867119871 and 119871119867) serve as a good candidatefor watermark embedding They are midfrequency subbandsthus they are less noticeable andmore robust than the 119871119871 and119867119867 subbands respectively
22 Carrier Selection Regardless of the insertion methodthere are always two ways to embed a watermark either to
insert one bit in a single coefficient or one bit in multiplecoefficients The latter solution leads to some importantadvantages of both robustness and imperceptibility [36]In fact a good watermark carrier has to guarantee twomain properties a high robustness combined with a lowperceptibility From a carrierrsquos point of view these prop-erties are manifested by a low distortion after attacks andan imperceptible change after embedding the watermarkThe blocksrsquo mean values (BMV) of 119867119871 and 119871119867 subbandssatisfy the above requirements it is computed from multiplecoefficients and it manifests a high robustness and imper-ceptibility To prove this claim we analyze the variation ofBMVs through the cumulative distribution function (CDF)where each block is constructed by gathering eight coef-ficients from unrepeated random positions The insertionof a watermark bit into a BMV can only be accomplishedthrough the modification of its block-coefficients Henceto emulate the insertion we propose to modify all thecoefficients of each block by a fixed value (negative orpositive)
First we investigate the quality of the resulting imageTheCDF in Figure 3(b) shows the large variations of the BMVsexpressed in terms of percentage the variation reaches upto 2 000 where 58 of the BMVs are modified between100 and 950 from their original values Despite this largevariation the modified image (Figure 2(b)) remains highlysimilar to the original one (Figure 2(a))
Concerning the robustness Figures 3(c) and 3(d) depictthe variation of BMVs always in terms of percentage afterJPEG compression and median filtering respectively Herewe can observe the limited variation which implies a highrobustness In fact in 72 of the cases the variation is lessthan 10 after compression and less than 20 in 73 of thetimes after filtering operations In both cases the variation isless than 30 in 83 of the BMVs
4 Security and Communication Networks
(a) (b)
Figure 2 Lena image (a) original and (b) after modification of all the1198671198711 and 1198711198671 coefficients
0010203040506070809
1
CDF
|BMV|
0 10 20 30 40 50 60 70 80
(a)
0 2000400 800 1200 1600BMV variation ()
0
01
02
03
04
05
06
07
08
09CD
F
(b)
BMV variation ()
0
01
02
03
04
05
06
07
08
09
CDF
0 10 20 30 40 50 60 70 80 90 100
(c)
0
01
02
03
04
05
06
07
08
09
BMV variation ()
CDF
0 10 20 30 40 50 60 70 80 90 100
(d)
Figure 3 Absolute value of Cumulative histogram of (a) original BMVs (b) difference between the original BMVs and the altered ones ((a)and (b)) (c) difference after JPEG(30) compression and (d) difference after median filtering (3 times 3)
It is worth noting that despite the robustness and imper-ceptibility of the BMVs the final results will also depend onthe insertion and detection method described in the nextsection
3 The Proposed Scheme
This section describes the insertion and extraction processesof a binary watermark 119872 = 1198871 1198872 119887119898 in a grayscale
Security and Communication Networks 5
2timesℎ
2 times w
LHl
HLl
c1 c2 c3 cn
c1 c2 c3 cn
x0
b1 b2 b3 bm
Original Image
Host coefficientsShuffling Chaotic sequence
generation
Block partitioning
Block meanmodulation
Watermark bits
Reshuffling
Watermarked subbands
Watermarked coefficients
BL1 BL2BLlfloorntrfloor
BL1 BL
2 BL
m
BLm+1BLlfloorntrfloor
LHl
HLl
c1c2
c3cn
Figure 4 Block diagram of the watermark embedding scheme
image I with 2 times ℎ height and 2 times 119908 width At this point theembedding parameters are not specified but kept generalizedto point out the flexibility of the scheme At first the originalimage goes through 119897 levels of DWT in order to manifest thedifferent frequencies of the image For the reasons presentedin Section 2 the 119871119867119897 and the 119867119871 119897 subbands are selected forembedding Let 119871119867119897 = 119871119867(119894 119895) and 119867119871 119897 = 119867119871(119894 119895) bethe full set of carrier coefficients where 119894 isin [1 1199082119897minus1] and119895 isin [1 ℎ2119897minus1] The coefficients of the two subbands are thenconcatenated into a single vector 119862 = 119888119896 where 119896 isin [1 119899]and 119899 = ℎ times 11990822119897minus3 Next the carrier vector 119862 is shuffledusing a chaotic sequence generated using the logistic mapand eachwatermark bit is embedded bymodulating themeanvalue of a set of coefficients The complete block diagramof the embedding process is given in Figure 4 The threemain stages involved in the watermarking process (chaoticsequence generation coefficients shuffling and mean blockmodulation) are detailed in the rest of this section
31 Chaotic Sequence Generation Chaotic maps are dynamicsystems defined by a mathematical function ((119909119899+1 = 119891(119909119899))that model a chaotic behavior In practice a variety of chaoticsystems may be utilized (eg Bakerrsquos map circle map duffingmap logistic map etc) The implementation of the latter isone of the simplest thanks to its polynomial mapping degreethough itrsquos high dynamic complexity These features allowedthe logistic map to attract a lot of attention in the informationsecurity disciplines such as cryptography and watermarking
The logistic map may be described by the followingrecurrence relation
119909119899+1 = 119891 (119909119899) = 120582 (119909119899 minus 1199092119899) (1)
The chaotic sequence 119909119899 generated by the function 119891 after119899 iterations is determined by its input parameters 120582 and 1199090where 1199090 is the initial value of the sequence and 120582 is thecontrol parameters In practice 1199090 is choosing between [0 1]and120582 lies between 3569955672 and 4These ranges guaranteea chaotic sequence with a high randomness in the outputwhere the slightest difference in the initial parameters (1199090 and120582) will lead to uncorrelated sequences over time
32 Shuffling In the shuffling phase the elements of the vec-tor 119862 are assigned to new positions according the generatedchaotic sequence The objective of this shuffling is double-fold Firstly it aims to improve the security of the schemethe 119905 coefficients forming each block (119861119871) are randomlyselected therebymaking it difficult for a third party to extractthe watermark or selectively attack the host coefficients todestroy the watermark without the permutation sequenceSecondly the shuffling dispels the correlation between adja-cent coefficients In fact adjacent coefficients tend to beof the same sign leading to MVBs distant from zero Incontrast the MVBs of scattered coefficients are gatheredaround the zero value which will serve us to minimizethe embedding strength and hence the perceptibility of thewatermark
The logistic map defined by (1) is employed to producea pseudo-random key containing the new positions for eachcoefficient
Let 119909119899 = (1199091 1199092 1199093 119909119899) be the chaotic sequencegenerated by means of the logistic map using the secretkey SK = 1199090 120582 where 119899 is the number of iterationsthat is equal to the total number of coefficients in 119871119867119897 and119867119871 119897 The 119909119899 coefficients are then sorted in a ascending
6 Security and Communication Networks
LL
HH
12
Chaotic sequence
4 2 7 1 6 5 3 8
12 09 -03 01 -24 -11 -07 -18
01 09 -07 12 -11 -24 -03 -18
022 025 028 067 074 081 090 0914 2 7 1 6 5 3 8
067 025 090 022 081 074 028 0911 2 3 4 5 6 7 8
Permutation sequence
Sorting09
-03 01
-24 -11
-07 -18C
C
Figure 5 Shuffling of DWT diagonal subbands using a chaotic sequence of eight real numbers
order (1199091199041 1199091199042 1199091199043 119909119904119899) and so the permutation sequence(1199041 1199042 1199043 119904119899) is obtained
The example shown in Figure 5 illustrates a concreteexample for the shuffling operation In this example 119899 is equalto eight the vector 119862 is the concatenation of the diagonalDWT coefficients and 119862 is the shuffled version of 11986233 Block Mean Modulation The first step in the blockmodulation phase is the partitioning of the confused hostcoefficients 119888119894 into independent blocks This operation isrealized by gathering each 119905 adjacent coefficients into a singleblock 119861119871 119894 as given by (2) where 119894 is the index of the blockand is the concatenation symbol The total number ofgenerated blocks 119861119871 should be larger or equal than the binarywatermark message length In other words lfloor119899119905rfloor ge 119898 suchas 119905 isin Nlowast
To embed a robust watermark we consider the quantizationtechnique which modulates the carrier hosts to 119875 nonover-lapping sets where 119875 represents the possible embeddedwords In this work the watermark is a binary messagehence the carrier will be modulated to two different sets(119901 = 2) each associated to a watermark bit The robustnessof the quantization-based insertion method depends on thedistance between the two sets On the other hand wideningthe gap will lead tomediocre results in terms of visual qualityTo work around this issue Chen et al [10] proposed toincrease the element of each set and it is up to the embeddingalgorithm to choose the appropriate quantizer among theavailable ones However as illustrated in Figure 3(a) mostof the BMV are concentrated around the zero where in 80of the BMVs are in the range [minus20 20] Thus we propose tomodulate the blocksrsquo mean values by a positive or negativenumber according to the embedded bit Let 119861119871 119894 be the meanvalue of the block 119894 defined by
119861119871 119894 = 1119905119894times119905sum
119895=119905(119894minus1)+1
119888119895 (3)
The modulation of 119861119871 119894 goes by firstly nullifying the meanvalue by spreading its value equally to all blockrsquos coefficientsthen adding or subtracting a threshold 119879 according to theembedded bit 119887 The equation of modulation is given by (4)where 119888119896 is a coefficient belonging to the block 119894 (119861119871 119894)
1198881015840119896 = 119888119896 minus 119861119871 119894119905 minus 119879 if 119887 = 0119888119896 minus 119861119871 119894119905 + 119879 otherwise (4)
Once all the watermark bits are inserted the watermarkedcoefficients are rearranged into their initial positions usingthe same key SK Finally the inverse DWT is performed toobtain the watermarked image
34 Watermark Extraction In the watermark extractionphase the watermarked image is firstly transformed in thefrequency domain as in the embedding and the same key SKis used to constitute different blocks The decision upon theembedded bit is judged according to the blockrsquos mean valueof the attacked watermarked block 119861119871119894
119887 = 1 if 119861119871119894 gt 00 otherwise (5)
4 Experimental Results
This section is dedicated to evaluate the performance ofthe suggested watermarking framework A variety of 8-bitdepth (119889 = 8) grayscale standard benchmark images witha resolution of 512 times 512 pixels (119908 = ℎ = 256) have beensubjects of assessment The used image database includesLena Baboon Peppers and Tank These images present avariety of image properties (eg mixture of dark and lightarea alongside with highly textures regions (hair) in the Lenaimage) which refutes the idea of tying the performance of thescheme to some image properties
The performance of the suggested method is experimen-tally examined from different aspects Firstly the relation
Security and Communication Networks 7
38 41 45PSNR (dB)
2
4
6
8
10
12
14
16
18
Thre
shol
d va
lue (
T)
123
DWT level
Figure 6 Relation of threshold value (T) and the PSNR at different DWT levels
between the capacity and the imperceptibility is evaluatedusing different insertion parameters Next the evolution ofcomplexity according to the host image size is monitoredAfterwards the robustness is validated in two ways initiallythe robustness against different levels of attacks is examinedand then a comparative study with similar works is presentedFinally the security of the scheme is verified
41 Imperceptibility The imperceptibility feature reflects thefidelity of the watermarked image toward the original oneSeveral metrics can be used to quantify the noticeable differ-ence between two images involving subjective and objectivemetrics In this work we employed the Peek Signal to NoiseRatio (PSNR) and the Structural Similarity Index Measure(SSIM) given by (6) and (7) respectively
PSNR (I Ilowast) = 10 log10( (2119889 minus 1)2 times 119908 times ℎsum119908119894=1sumℎ119894=1 (I119894119895 minus Ilowast119894119895)2) (6)
where 120579I and 120583I are respectively the mean intensity and thestandard deviation of the original image I and 120590119868119868lowast is thesample cross-correlation between I and Ilowast The constants 1198781and 1198782 are defined as 1198781 = (1198961(2119889minus1))2 and 1198782 = (1198962(2119889minus1))2where 1198961 1198962 ≪ 142 Capacity The maximum capacity of the proposedscheme depends on three variables the image dimension (119908
and ℎ) the wavelet level (119897) and the block length (119905) Hencewe can express the maximum capacity using
Max capacity = lfloor 119908 times ℎ119905 times 22119897minus3rfloor (8)
For a better understanding of the relation between the capac-ity and imperceptibility we embedded a variable message size(119898 isin 256 512 1024) at different levels of decomposition(119897 isin 1 2 3) into three different images namely Lena Coupleand Baboon For each watermarking operation the blocksrsquosize 119905 is maximized so that all coefficients are involved inthe watermarking process (eg 119897 = 2 and 119905 = 32 for119898 = 1024) The threshold 119879 is manually chosen to reach aPSNR value of 38 41 and 45 dBThe parameters of insertion(119879 119905 119897 119898) the objective measurements (PSNR and SSIM)and the watermarked images are all listed in Table 1
The first observation that we draw from Table 1 is thatfor a given PSNR the threshold 119879 is converging towardthe same value regardless the host image or the watermarksize This amounts to saying that we can anticipate the119879 value according to the level of transformation and thetargeted PSNR Figure 6 illustrates the relation between theseparameters where119879 is obtained by averaging all the values fora specific level and PSNR Secondly it is shown from Table 1that at 45 dB the distortion due to the watermark insertionis invisible whatever the size of the message and the levelof decomposition However for watermarked images with aPSNR of 38 and 41 dB the imperceptibility is conditionedupon the level 119897 the higher 119897 is the lower the quality is goingto be
43 Complexity The computational cost of the watermarkembedding and extraction is an important aspect that may
8 Security and Communication Networks
Table 1 Quality of watermarked images using different insertion parameters (119897 119879119898)Message size Transformation
level Watermarked Lena images Watermarked Coupleimages
Figure 7 Computational time of (a) overall insertion and detection and (b) details of the embedding stage
restrict or expand the use of a given watermarking schemeIn the literature most of schemes are evaluated in a standardimage with a size of 512 times 512 pixels and there is noperspective of the computational cost trend when the imagegoes bigger which is more realistic Hence we propose tovary the size of the image beginning with 64 times 64 pixelsand ending with 4096 times 4096 pixels by doubling the heightand width after each insertion For the sake of simplicity thewatermark was embedded after one level of DWT (119897 = 1)with a fixed block size (119905 = 4) and a maximum watermarksize as defined by (8)
Figure 7(a) depicts the execution time evolution ofthe overall embedding and detection alongside with thevariation of the watermark size (red) as a function of theimage size The details of the embedding phase are shownin Figure 7(b) The most important remark to be made fromFigure 7 is that the computational time evolves broadly inline with the watermark size (red curve) to find out thecomplexity of our scheme in comparison with other state-ofthe-art methods
44 Robustness In the following we investigate the robust-ness of the suggested watermarking approach under severaltypes of attacks The robustness is quantified using the BitError Rate (BER) and the Normalized Correlation (NC)metrics
441 Evaluated Attacks Since the proposed scheme isdesigned to be robust the embedded message should lastor at list a significant part of it after image processingoperations Fromwatermark perspective all image processingoperations whatever performed on goodwill or not is consid-ered as attacks The involved attacks in the evaluation tests areclassified into three groups
(i) Compression attacks usually images are not storedin a raw format but after a compression operationthat aims to reduce the required storage space JPEGalgorithm is one of the most used compression
algorithms However JPEG is a lossy algorithm thatcauses some quality degradation inversely commen-surate with the compression quality factor Hence wetested the robustness of our algorithm under differentJPEG quality factors ranging from 10 to 90
(ii) Image processing attacks used to imitate a noisecontaminating the image during transmission such assalt and peppers or to improve the imagersquos qualityby eliminating a noise or enhancing some propertiesThe used attacks are listed as follows scaling medianfiltering sharpening Gaussian filtering average fil-tering Gaussian noise salt and pepper noise contam-ination histogram equalization and additive whiteGaussian noise (AWGN)
(iii) Geometric attacks used to correct geometric distor-tions such camera orientation (rotation) or imagesrsquoalignment (cropping) The used attacks are listed asfollows image cropping and resizing and rotation
Since the DWT coefficients are not geometrically invari-ant a manual synchronization of the imagersquos dimension andorientation was realized prior to the extraction
442 Performance Evaluation Here we test the robustnessof our scheme on three different images Boat Peppersand Barbara Each time we varied the embedded messagebetween 128 bits and 2048 bits The embedding threshold 119879is manually tuned to provide a watermarked image with aPSNR value of 40 dB and 45 dB The robustness is expressedin terms of BER () after mean filter (Figure 8) JPEGcompression (Figure 9) median filter (Figure 10) and gammacorrection Poisson noise histogram equalization and imageunsharpening (Figure 11) Figures 8ndash11 show that the robust-ness performance slightly differs between images For a smallmessage length (between 128 and 256 bits) and a PSNR valueof 40 dB the BER does not exceed 20 for most of attacksHowever the larger the message size is or the stronger theattack is the higher the BER will be
10 Security and Communication Networks
128 256 512 1024 2048Watermark size
0
10
20
30
40
50BE
R (
)
Mean filter (3x3)Mean filter (5x5)Mean filter (7x7)Mean filter (9x9)BarbaraPeppersBoat
(a)
Mean filter (3x3)Mean filter (5x5)Mean filter (7x7)Mean filter (9x9)BarbaraPeppersBoat
128 256 512 1024 2048Watermark size
0
10
20
30
40
50
BER
()
(b)
Figure 8 BER() after mean filter attacks using different message lengths at different PSNR values (a) 40 dB and (b) 45 dB
Figure 9 BER() after JPEG compression using different message lengths at different PSNR values (a) 40 dB and (b) 45 dB
443 Comparative Study To prove the superiority and therobustness advantages of our scheme a comparative studywith state of the art works is indispensable In fact wecompared the robustness of our scheme with six robustwatermarking schemes [26 27 35 37ndash39] In order to carryout a fair comparison the same insertion parameters (PSNRwatermark size) and attacks are used in this comparisonThe PSNRs value of different images and their correspondingvalues in the comparative schemes are illustrated in Figure 12It is worth mentioning that the insertion parameters (119897 119905 119879)were empirically chosen to produce the same PSNR alongsidewith an acceptable visual quality
The robustness results are presented in Tables 2ndash5 interms of either BER or NC depending on how they appearin the reference papers and the superior results are markedin bold number
Based on the simulation results given in Tables 2ndash5 it canbe found that our approach outperforms similar schemes on7194 of the time Equal results are obtained on 2343 ofthe time and our scheme underperforms on 462 of thecases
45 Security The robustness of a watermarking scheme isnot sufficient to judge a scheme as reliable As discussedin the first section robust watermarking is used to secureimages Hence the security is necessary First the watermarkshould be immune against illegal extraction even when theembedding procedure is exposed Second the watermarkdetector should not report the existence of the watermarkwhen none is embedded To evaluate these two risks weextract the watermark from a watermarked image using
Security and Communication Networks 11
Table 2 BER comparison with [27] and [37] using a message length of 128 bits
different keys and extract a watermark from a nonwater-marked image respectively In the two cases the maximumcorrelation obtained has not exceeded the value 059 whichpoints out a negative presence
5 Conclusion
The mean value of DWTrsquos vertical and horizontal subbandscoefficients is highly robust and imperceptible In fact these
two advantages were exploited in the design of a robustand blind watermarking approach The embedding methodconsisted in modulating the BMV in accordance with thequantization technique The use of a single transformationdomain alongside with a simple embedding technique makesfrom the proposed algorithmcomputationally undemandingAt the extraction phase the watermark bits were identifiedby the BMV position with respect to the zero value Unlikesome existing schemes the extraction rules did not require to
Security and Communication Networks 15
be adjusted to the applied attacks to fulfill a high robustnessIn fact the inserted watermark was capable to withstand withno error after JPEG compression up to a quality factor of 30Gaussian filter and other attacks
Data Availability
The data used to support the findings of this study areincluded within the article
Conflicts of Interest
The authors declare that they have no conflicts of interest
References
[1] S Liansheng Z Bei W Zhanmin and T Ailing ldquoAn opticalcolor image watermarking scheme by using compressive sens-ingwith human visual characteristics in gyrator domainrdquoOpticsand Lasers in Engineering vol 92 pp 85ndash93 2017
[2] D Shukla and M Sharma ldquoA Novel Scene-Based Video Water-marking Scheme for Copyright Protectionrdquo Journal of IntelligentSystems vol 27 no 1 pp 47ndash66 2018
[3] A Tiwari and M Sharma ldquoAn efficient vector quantizationbased watermarking method for image integrity authentica-tionrdquo Advances in Intelligent Systems and Computing vol 519pp 215ndash225 2018
[4] S Bravo-Solorio F Calderon C-T Li and A K Nandi ldquoFastfragile watermark embedding and iterative mechanism withhigh self-restoration performancerdquo Digital Signal Processingvol 73 pp 83ndash92 2018
[5] A Ghardallou A Kricha A Sakly and A Mtibaa ldquoAdaptiveblock sized reversible watermarking scheme based on integertransformrdquo in Proceedings of the 17th International Conferenceon Sciences and Techniques of Automatic Control and ComputerEngineering STA 2016 pp 347ndash351 Tunisia December 2016
[6] R G van Schyndel A Z Tirkel and C F Osborne ldquoA digitalwatermarkrdquo in Proceedings of the IEEE International ConferenceImage Processing (ICIP rsquo94) vol 2 pp 86ndash90 Austin Tex USANovember 1994
[7] W-H Lin S-J Horng T-W Kao P Fan C-L Lee and Y PanldquoAn efficient watermarking method based on significant differ-ence of wavelet coefficient quantizationrdquo IEEE Transactions onMultimedia vol 10 no 5 pp 746ndash757 2008
[8] S Bhattacharyya and G Sanyal ldquoA Robust Image Steganogra-phy using DWT Difference Modulation (DWTDM)rdquo Interna-tional Journal of Computer Network and Information Securityvol 4 no 7 pp 27ndash40 2012
[9] H S Malvar and D A Florencio ldquoImproved spread spectruma new modulation technique for robust watermarkingrdquo IEEETransactions on Signal Processing vol 51 no 4 pp 898ndash9052003
[10] B Chen and G W Wornell ldquoProvably robust digital water-markingrdquo in Proceedings of the 1999 Multimedia Systems andApplications II pp 43ndash54 September 1999
[11] Z Kricha A Kricha and A Sakly ldquoAccommodative extractorfor QIM-based watermarking schemesrdquo IET Image Processing2018
[12] I J Cox J Kilian T Leighton and T Shamoon ldquoSecure imper-ceptable yet perceptually salient spread spectrum watermark
formultimediardquo in Proceedings of the 1996 Southcon Conferencepp 192ndash197 June 1996
[13] X Li J Liu J Sun Z Sun H Hu and L Zhang ldquoImprovedspread transform dither modulation-based watermark algo-rithm based on step projectionrdquo in Proceedings of the 2010 IEEEInternational Symposium on Broadband Multimedia Systemsand Broadcasting (BMSB) pp 1ndash5 Shanghai China March2010
[14] H-T Hu and L-Y Hsu ldquoExploring DWT-SVD-DCT featureparameters for robust multiple watermarking against JPEG andJPEG2000 compressionrdquo Computers and Electrical Engineeringvol 41 pp 52ndash63 2015
[15] J Cannons and P Moulin ldquoDesign and statistical analysis of ahash-aided image watermarking systemrdquo IEEE Transactions onImage Processing vol 13 no 10 pp 1393ndash1408 2004
[16] A Valizadeh and Z J Wang ldquoAn improved multiplicativespread spectrum embedding scheme for data hidingrdquo IEEETransactions on Information Forensics and Security vol 7 no4 pp 1127ndash1143 2012
[17] M Kutter D Frederic and F B JordanDigital signature of colorimages using amplitude modulation 1997
[18] Q Su and B Chen ldquoRobust color image watermarking tech-nique in the spatial domainrdquo Soft Computing vol 22 no 1 pp91ndash106 2018
[19] Z Yanling Z Xiaoshi L Na et al ldquoA Digital ImageWatermarkAlgorithm Based onDCCoefficientsQuantizationrdquo in Proceed-ings of the 2006 6th World Congress on Intelligent Control andAutomation pp 9734ndash9738 Dalian China 2006
[20] RMZhaoH LianHWPang andBNHu ldquoAWatermarkingAlgorithm by Modifying AC Coefficies in DCT Domainrdquo inProceedings of the 2008 International Symposium on InformationScience and Engineering (ISISE) pp 159ndash162 Shanghai ChinaDecember 2008
[21] B L Gunjal and R RManthalkar ldquoDiscreteWavelet Transformbased strongly robust watermarking scheme for informationhiding in digital imagesrdquo in Proceedings of the 3rd InternationalConference on Emerging Trends in Engineering and TechnologyICETET 2010 pp 124ndash129 India November 2010
[22] H Shi and F Lv ldquoA Blind Digital Watermark Techniquefor Color Image Based on Integer Wavelet Transformrdquo inProceedings of the 2010 International Conference on BiomedicalEngineering and Computer Science (ICBECS) pp 1ndash4 WuhanChina April 2010
[23] S Ranjbar F Zargari and M Ghanbari ldquoA highly robust two-stage Contourlet-based digital image watermarking methodrdquoSignal Processing Image Communication vol 28 no 10 pp1526ndash1536 2013
[24] Q Su and B Chen ldquoA novel blind color image watermarkingusing upper Hessenbergmatrixrdquo AEU - International Journal ofElectronics and Communications vol 78 pp 64ndash71 2017
[25] F Ourique V Licks R Jordan and F Perez-Gonzalez ldquoAngleQIM a novel watermark embedding scheme robust againstamplitude scaling distortionsrdquo in Proceedings of the IEEE Inter-national Conference on Acoustics Speech and Signal Processing(ICASSP rsquo05) pp 797ndash800 March 2005
[26] H R Fazlali S Samavi N Karimi and S Shirani ldquoAdaptiveblind image watermarking using edge pixel concentrationrdquoMultimedia Tools and Applications vol 76 no 2 pp 3105ndash31202017
[27] X-Y Wang Y-N Liu H Xu A-L Wang and H-Y YangldquoBlind optimum detector for robust image watermarking in
16 Security and Communication Networks
nonsubsampled shearlet Domainrdquo Information Sciences vol372 pp 634ndash654 2016
[28] X Liu C Lin and S Yuan ldquoBlind Dual Watermarking forColor Imagesrsquo Authentication and Copyright Protectionrdquo IEEETransactions on Circuits and Systems for Video Technology vol28 no 5 pp 1047ndash1055 2018
[29] Poonam and S M Arora ldquoA DWT-SVD based Robust DigitalWatermarking for Digital Imagesrdquo Procedia Computer Sciencevol 132 pp 1441ndash1448 2018
[30] S Fazli and M Moeini ldquoA robust image watermarking methodbased on DWT DCT and SVD using a new technique forcorrection of main geometric attacksrdquo Optik - InternationalJournal for Light and ElectronOptics vol 127 no 2 pp 964ndash9722016
[31] K Kuppusamy and K Thamodaran ldquoOptimized Image Water-marking Scheme Based On PSOrdquo Procedia Engineering vol 38pp 493ndash503 2012
[32] V S Rao R S Shekhawat and V K Srivastava ldquoA dwt-dct-svd based digital image watermarking scheme using particleswarm optimizationrdquo in Proceedings of the 2012 IEEE StudentsrsquoConference on Electrical Electronics and Computer Science pp1ndash4 2012
[33] T T Takore P R Kumar and G L Devi ldquoA modified blindimage watermarking scheme based on DWT DCT and SVDdomain using GA to optimize robustnessrdquo in Proceedings ofthe 2016 International Conference on Electrical Electronics andOptimization Techniques ICEEOT 2016 pp 2725ndash2729 IndiaMarch 2016
[34] H Hu S Lin and L Hsu ldquoEffective blind speech watermarkingvia adaptive mean modulation and package synchronization inDWT domainrdquo EURASIP Journal on Audio Speech and MusicProcessing vol 2017 no 1 2017
[35] W-H Lin Y-R Wang S-J Horng T-W Kao and Y Pan ldquoAblind watermarking method using maximum wavelet coeffi-cient quantizationrdquo Expert Systems with Applications vol 36no 9 pp 11509ndash11516 2009
[36] X Li J Liu J Sun X Yang andW Liu ldquoStep-projection-basedspread transform dither modulationrdquo IET Information Securityvol 5 no 3 pp 170ndash180 2011
[37] M Hamghalam S Mirzakuchaki andM A Akhaee ldquoGeomet-ric modelling of the wavelet coefficients for image watermark-ing using optimum detectorrdquo IET Image Processing vol 8 no 3pp 162ndash172 2014
[38] V S Verma R K Jha and A Ojha ldquoDigital watermarkextraction using support vector machine with principal com-ponent analysis based feature reductionrdquo Journal of VisualCommunication and Image Representation vol 31 pp 75ndash852015
[39] C Li Z Zhang Y Wang B Ma and D Huang ldquoDithermodulation of significant amplitude difference for waveletbased robust watermarkingrdquo Neurocomputing vol 166 pp404ndash415 2015
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Navigation and Observation
International Journal of
Hindawi
wwwhindawicom Volume 2018
Advances in
Multimedia
Submit your manuscripts atwwwhindawicom
Security and Communication Networks 3
64 128
256
512
1024
2048
4096
Image heightwidth (pixels)
0
10
20
30
40
50
60
70
80
90
Com
puta
tion
time (
s)
DWTDCTShearlet
2D Transformation
Figure 1 Computational time evolution with respect to the image dimension of some 2D transformations
In order to prove the complexity advantage of the DWTover other transformations the computational time of theDWT(1 level) DCT (8times 8 blocks) and the Shearlet transformare examined using different image sizes (Figure 1) Thesetransformations are employed in [7 28 35] [13 20 36]and [27] respectively The advantage of the DWT over othertransformations is clear for large image sizes and especiallyover the DCT which is the most time consuming among thethree evaluated transformations
Due to the above advantages we adopted the DWTtransformation in our schemeWhen applied to an image theDWT is achieved by filtering separately the rows and columnsof the image Each individual image transformation producesfour separable subbands one coarse subimage locating thelow frequencies (119871119871) and three locating the horizontal (119867119871)vertical (119871119867) and diagonal (119871119871) details (high frequencies)The WT may be repeated as appropriate by filtering againthe low frequency subband Each individual operation stageproduce a resolution level and each higher level yields to abetter frequency localization but a shrinkage in the subbandsize From the perspective of imperceptibility the119871119871 subbandis inappropriate for watermark insertion The approximationcoefficients represent the low frequencies that reflect signif-icant information about an image Thus the manipulationof these coefficients provokes quality degradation On theother hand the modification of the 119867119867 coefficients maygo unnoticed but the watermark may be easily removedby image processing tasks such as lossy compression Theremaining subbands (119867119871 and 119871119867) serve as a good candidatefor watermark embedding They are midfrequency subbandsthus they are less noticeable andmore robust than the 119871119871 and119867119867 subbands respectively
22 Carrier Selection Regardless of the insertion methodthere are always two ways to embed a watermark either to
insert one bit in a single coefficient or one bit in multiplecoefficients The latter solution leads to some importantadvantages of both robustness and imperceptibility [36]In fact a good watermark carrier has to guarantee twomain properties a high robustness combined with a lowperceptibility From a carrierrsquos point of view these prop-erties are manifested by a low distortion after attacks andan imperceptible change after embedding the watermarkThe blocksrsquo mean values (BMV) of 119867119871 and 119871119867 subbandssatisfy the above requirements it is computed from multiplecoefficients and it manifests a high robustness and imper-ceptibility To prove this claim we analyze the variation ofBMVs through the cumulative distribution function (CDF)where each block is constructed by gathering eight coef-ficients from unrepeated random positions The insertionof a watermark bit into a BMV can only be accomplishedthrough the modification of its block-coefficients Henceto emulate the insertion we propose to modify all thecoefficients of each block by a fixed value (negative orpositive)
First we investigate the quality of the resulting imageTheCDF in Figure 3(b) shows the large variations of the BMVsexpressed in terms of percentage the variation reaches upto 2 000 where 58 of the BMVs are modified between100 and 950 from their original values Despite this largevariation the modified image (Figure 2(b)) remains highlysimilar to the original one (Figure 2(a))
Concerning the robustness Figures 3(c) and 3(d) depictthe variation of BMVs always in terms of percentage afterJPEG compression and median filtering respectively Herewe can observe the limited variation which implies a highrobustness In fact in 72 of the cases the variation is lessthan 10 after compression and less than 20 in 73 of thetimes after filtering operations In both cases the variation isless than 30 in 83 of the BMVs
4 Security and Communication Networks
(a) (b)
Figure 2 Lena image (a) original and (b) after modification of all the1198671198711 and 1198711198671 coefficients
0010203040506070809
1
CDF
|BMV|
0 10 20 30 40 50 60 70 80
(a)
0 2000400 800 1200 1600BMV variation ()
0
01
02
03
04
05
06
07
08
09CD
F
(b)
BMV variation ()
0
01
02
03
04
05
06
07
08
09
CDF
0 10 20 30 40 50 60 70 80 90 100
(c)
0
01
02
03
04
05
06
07
08
09
BMV variation ()
CDF
0 10 20 30 40 50 60 70 80 90 100
(d)
Figure 3 Absolute value of Cumulative histogram of (a) original BMVs (b) difference between the original BMVs and the altered ones ((a)and (b)) (c) difference after JPEG(30) compression and (d) difference after median filtering (3 times 3)
It is worth noting that despite the robustness and imper-ceptibility of the BMVs the final results will also depend onthe insertion and detection method described in the nextsection
3 The Proposed Scheme
This section describes the insertion and extraction processesof a binary watermark 119872 = 1198871 1198872 119887119898 in a grayscale
Security and Communication Networks 5
2timesℎ
2 times w
LHl
HLl
c1 c2 c3 cn
c1 c2 c3 cn
x0
b1 b2 b3 bm
Original Image
Host coefficientsShuffling Chaotic sequence
generation
Block partitioning
Block meanmodulation
Watermark bits
Reshuffling
Watermarked subbands
Watermarked coefficients
BL1 BL2BLlfloorntrfloor
BL1 BL
2 BL
m
BLm+1BLlfloorntrfloor
LHl
HLl
c1c2
c3cn
Figure 4 Block diagram of the watermark embedding scheme
image I with 2 times ℎ height and 2 times 119908 width At this point theembedding parameters are not specified but kept generalizedto point out the flexibility of the scheme At first the originalimage goes through 119897 levels of DWT in order to manifest thedifferent frequencies of the image For the reasons presentedin Section 2 the 119871119867119897 and the 119867119871 119897 subbands are selected forembedding Let 119871119867119897 = 119871119867(119894 119895) and 119867119871 119897 = 119867119871(119894 119895) bethe full set of carrier coefficients where 119894 isin [1 1199082119897minus1] and119895 isin [1 ℎ2119897minus1] The coefficients of the two subbands are thenconcatenated into a single vector 119862 = 119888119896 where 119896 isin [1 119899]and 119899 = ℎ times 11990822119897minus3 Next the carrier vector 119862 is shuffledusing a chaotic sequence generated using the logistic mapand eachwatermark bit is embedded bymodulating themeanvalue of a set of coefficients The complete block diagramof the embedding process is given in Figure 4 The threemain stages involved in the watermarking process (chaoticsequence generation coefficients shuffling and mean blockmodulation) are detailed in the rest of this section
31 Chaotic Sequence Generation Chaotic maps are dynamicsystems defined by a mathematical function ((119909119899+1 = 119891(119909119899))that model a chaotic behavior In practice a variety of chaoticsystems may be utilized (eg Bakerrsquos map circle map duffingmap logistic map etc) The implementation of the latter isone of the simplest thanks to its polynomial mapping degreethough itrsquos high dynamic complexity These features allowedthe logistic map to attract a lot of attention in the informationsecurity disciplines such as cryptography and watermarking
The logistic map may be described by the followingrecurrence relation
119909119899+1 = 119891 (119909119899) = 120582 (119909119899 minus 1199092119899) (1)
The chaotic sequence 119909119899 generated by the function 119891 after119899 iterations is determined by its input parameters 120582 and 1199090where 1199090 is the initial value of the sequence and 120582 is thecontrol parameters In practice 1199090 is choosing between [0 1]and120582 lies between 3569955672 and 4These ranges guaranteea chaotic sequence with a high randomness in the outputwhere the slightest difference in the initial parameters (1199090 and120582) will lead to uncorrelated sequences over time
32 Shuffling In the shuffling phase the elements of the vec-tor 119862 are assigned to new positions according the generatedchaotic sequence The objective of this shuffling is double-fold Firstly it aims to improve the security of the schemethe 119905 coefficients forming each block (119861119871) are randomlyselected therebymaking it difficult for a third party to extractthe watermark or selectively attack the host coefficients todestroy the watermark without the permutation sequenceSecondly the shuffling dispels the correlation between adja-cent coefficients In fact adjacent coefficients tend to beof the same sign leading to MVBs distant from zero Incontrast the MVBs of scattered coefficients are gatheredaround the zero value which will serve us to minimizethe embedding strength and hence the perceptibility of thewatermark
The logistic map defined by (1) is employed to producea pseudo-random key containing the new positions for eachcoefficient
Let 119909119899 = (1199091 1199092 1199093 119909119899) be the chaotic sequencegenerated by means of the logistic map using the secretkey SK = 1199090 120582 where 119899 is the number of iterationsthat is equal to the total number of coefficients in 119871119867119897 and119867119871 119897 The 119909119899 coefficients are then sorted in a ascending
6 Security and Communication Networks
LL
HH
12
Chaotic sequence
4 2 7 1 6 5 3 8
12 09 -03 01 -24 -11 -07 -18
01 09 -07 12 -11 -24 -03 -18
022 025 028 067 074 081 090 0914 2 7 1 6 5 3 8
067 025 090 022 081 074 028 0911 2 3 4 5 6 7 8
Permutation sequence
Sorting09
-03 01
-24 -11
-07 -18C
C
Figure 5 Shuffling of DWT diagonal subbands using a chaotic sequence of eight real numbers
order (1199091199041 1199091199042 1199091199043 119909119904119899) and so the permutation sequence(1199041 1199042 1199043 119904119899) is obtained
The example shown in Figure 5 illustrates a concreteexample for the shuffling operation In this example 119899 is equalto eight the vector 119862 is the concatenation of the diagonalDWT coefficients and 119862 is the shuffled version of 11986233 Block Mean Modulation The first step in the blockmodulation phase is the partitioning of the confused hostcoefficients 119888119894 into independent blocks This operation isrealized by gathering each 119905 adjacent coefficients into a singleblock 119861119871 119894 as given by (2) where 119894 is the index of the blockand is the concatenation symbol The total number ofgenerated blocks 119861119871 should be larger or equal than the binarywatermark message length In other words lfloor119899119905rfloor ge 119898 suchas 119905 isin Nlowast
To embed a robust watermark we consider the quantizationtechnique which modulates the carrier hosts to 119875 nonover-lapping sets where 119875 represents the possible embeddedwords In this work the watermark is a binary messagehence the carrier will be modulated to two different sets(119901 = 2) each associated to a watermark bit The robustnessof the quantization-based insertion method depends on thedistance between the two sets On the other hand wideningthe gap will lead tomediocre results in terms of visual qualityTo work around this issue Chen et al [10] proposed toincrease the element of each set and it is up to the embeddingalgorithm to choose the appropriate quantizer among theavailable ones However as illustrated in Figure 3(a) mostof the BMV are concentrated around the zero where in 80of the BMVs are in the range [minus20 20] Thus we propose tomodulate the blocksrsquo mean values by a positive or negativenumber according to the embedded bit Let 119861119871 119894 be the meanvalue of the block 119894 defined by
119861119871 119894 = 1119905119894times119905sum
119895=119905(119894minus1)+1
119888119895 (3)
The modulation of 119861119871 119894 goes by firstly nullifying the meanvalue by spreading its value equally to all blockrsquos coefficientsthen adding or subtracting a threshold 119879 according to theembedded bit 119887 The equation of modulation is given by (4)where 119888119896 is a coefficient belonging to the block 119894 (119861119871 119894)
1198881015840119896 = 119888119896 minus 119861119871 119894119905 minus 119879 if 119887 = 0119888119896 minus 119861119871 119894119905 + 119879 otherwise (4)
Once all the watermark bits are inserted the watermarkedcoefficients are rearranged into their initial positions usingthe same key SK Finally the inverse DWT is performed toobtain the watermarked image
34 Watermark Extraction In the watermark extractionphase the watermarked image is firstly transformed in thefrequency domain as in the embedding and the same key SKis used to constitute different blocks The decision upon theembedded bit is judged according to the blockrsquos mean valueof the attacked watermarked block 119861119871119894
119887 = 1 if 119861119871119894 gt 00 otherwise (5)
4 Experimental Results
This section is dedicated to evaluate the performance ofthe suggested watermarking framework A variety of 8-bitdepth (119889 = 8) grayscale standard benchmark images witha resolution of 512 times 512 pixels (119908 = ℎ = 256) have beensubjects of assessment The used image database includesLena Baboon Peppers and Tank These images present avariety of image properties (eg mixture of dark and lightarea alongside with highly textures regions (hair) in the Lenaimage) which refutes the idea of tying the performance of thescheme to some image properties
The performance of the suggested method is experimen-tally examined from different aspects Firstly the relation
Security and Communication Networks 7
38 41 45PSNR (dB)
2
4
6
8
10
12
14
16
18
Thre
shol
d va
lue (
T)
123
DWT level
Figure 6 Relation of threshold value (T) and the PSNR at different DWT levels
between the capacity and the imperceptibility is evaluatedusing different insertion parameters Next the evolution ofcomplexity according to the host image size is monitoredAfterwards the robustness is validated in two ways initiallythe robustness against different levels of attacks is examinedand then a comparative study with similar works is presentedFinally the security of the scheme is verified
41 Imperceptibility The imperceptibility feature reflects thefidelity of the watermarked image toward the original oneSeveral metrics can be used to quantify the noticeable differ-ence between two images involving subjective and objectivemetrics In this work we employed the Peek Signal to NoiseRatio (PSNR) and the Structural Similarity Index Measure(SSIM) given by (6) and (7) respectively
PSNR (I Ilowast) = 10 log10( (2119889 minus 1)2 times 119908 times ℎsum119908119894=1sumℎ119894=1 (I119894119895 minus Ilowast119894119895)2) (6)
where 120579I and 120583I are respectively the mean intensity and thestandard deviation of the original image I and 120590119868119868lowast is thesample cross-correlation between I and Ilowast The constants 1198781and 1198782 are defined as 1198781 = (1198961(2119889minus1))2 and 1198782 = (1198962(2119889minus1))2where 1198961 1198962 ≪ 142 Capacity The maximum capacity of the proposedscheme depends on three variables the image dimension (119908
and ℎ) the wavelet level (119897) and the block length (119905) Hencewe can express the maximum capacity using
Max capacity = lfloor 119908 times ℎ119905 times 22119897minus3rfloor (8)
For a better understanding of the relation between the capac-ity and imperceptibility we embedded a variable message size(119898 isin 256 512 1024) at different levels of decomposition(119897 isin 1 2 3) into three different images namely Lena Coupleand Baboon For each watermarking operation the blocksrsquosize 119905 is maximized so that all coefficients are involved inthe watermarking process (eg 119897 = 2 and 119905 = 32 for119898 = 1024) The threshold 119879 is manually chosen to reach aPSNR value of 38 41 and 45 dBThe parameters of insertion(119879 119905 119897 119898) the objective measurements (PSNR and SSIM)and the watermarked images are all listed in Table 1
The first observation that we draw from Table 1 is thatfor a given PSNR the threshold 119879 is converging towardthe same value regardless the host image or the watermarksize This amounts to saying that we can anticipate the119879 value according to the level of transformation and thetargeted PSNR Figure 6 illustrates the relation between theseparameters where119879 is obtained by averaging all the values fora specific level and PSNR Secondly it is shown from Table 1that at 45 dB the distortion due to the watermark insertionis invisible whatever the size of the message and the levelof decomposition However for watermarked images with aPSNR of 38 and 41 dB the imperceptibility is conditionedupon the level 119897 the higher 119897 is the lower the quality is goingto be
43 Complexity The computational cost of the watermarkembedding and extraction is an important aspect that may
8 Security and Communication Networks
Table 1 Quality of watermarked images using different insertion parameters (119897 119879119898)Message size Transformation
level Watermarked Lena images Watermarked Coupleimages
Figure 7 Computational time of (a) overall insertion and detection and (b) details of the embedding stage
restrict or expand the use of a given watermarking schemeIn the literature most of schemes are evaluated in a standardimage with a size of 512 times 512 pixels and there is noperspective of the computational cost trend when the imagegoes bigger which is more realistic Hence we propose tovary the size of the image beginning with 64 times 64 pixelsand ending with 4096 times 4096 pixels by doubling the heightand width after each insertion For the sake of simplicity thewatermark was embedded after one level of DWT (119897 = 1)with a fixed block size (119905 = 4) and a maximum watermarksize as defined by (8)
Figure 7(a) depicts the execution time evolution ofthe overall embedding and detection alongside with thevariation of the watermark size (red) as a function of theimage size The details of the embedding phase are shownin Figure 7(b) The most important remark to be made fromFigure 7 is that the computational time evolves broadly inline with the watermark size (red curve) to find out thecomplexity of our scheme in comparison with other state-ofthe-art methods
44 Robustness In the following we investigate the robust-ness of the suggested watermarking approach under severaltypes of attacks The robustness is quantified using the BitError Rate (BER) and the Normalized Correlation (NC)metrics
441 Evaluated Attacks Since the proposed scheme isdesigned to be robust the embedded message should lastor at list a significant part of it after image processingoperations Fromwatermark perspective all image processingoperations whatever performed on goodwill or not is consid-ered as attacks The involved attacks in the evaluation tests areclassified into three groups
(i) Compression attacks usually images are not storedin a raw format but after a compression operationthat aims to reduce the required storage space JPEGalgorithm is one of the most used compression
algorithms However JPEG is a lossy algorithm thatcauses some quality degradation inversely commen-surate with the compression quality factor Hence wetested the robustness of our algorithm under differentJPEG quality factors ranging from 10 to 90
(ii) Image processing attacks used to imitate a noisecontaminating the image during transmission such assalt and peppers or to improve the imagersquos qualityby eliminating a noise or enhancing some propertiesThe used attacks are listed as follows scaling medianfiltering sharpening Gaussian filtering average fil-tering Gaussian noise salt and pepper noise contam-ination histogram equalization and additive whiteGaussian noise (AWGN)
(iii) Geometric attacks used to correct geometric distor-tions such camera orientation (rotation) or imagesrsquoalignment (cropping) The used attacks are listed asfollows image cropping and resizing and rotation
Since the DWT coefficients are not geometrically invari-ant a manual synchronization of the imagersquos dimension andorientation was realized prior to the extraction
442 Performance Evaluation Here we test the robustnessof our scheme on three different images Boat Peppersand Barbara Each time we varied the embedded messagebetween 128 bits and 2048 bits The embedding threshold 119879is manually tuned to provide a watermarked image with aPSNR value of 40 dB and 45 dB The robustness is expressedin terms of BER () after mean filter (Figure 8) JPEGcompression (Figure 9) median filter (Figure 10) and gammacorrection Poisson noise histogram equalization and imageunsharpening (Figure 11) Figures 8ndash11 show that the robust-ness performance slightly differs between images For a smallmessage length (between 128 and 256 bits) and a PSNR valueof 40 dB the BER does not exceed 20 for most of attacksHowever the larger the message size is or the stronger theattack is the higher the BER will be
10 Security and Communication Networks
128 256 512 1024 2048Watermark size
0
10
20
30
40
50BE
R (
)
Mean filter (3x3)Mean filter (5x5)Mean filter (7x7)Mean filter (9x9)BarbaraPeppersBoat
(a)
Mean filter (3x3)Mean filter (5x5)Mean filter (7x7)Mean filter (9x9)BarbaraPeppersBoat
128 256 512 1024 2048Watermark size
0
10
20
30
40
50
BER
()
(b)
Figure 8 BER() after mean filter attacks using different message lengths at different PSNR values (a) 40 dB and (b) 45 dB
Figure 9 BER() after JPEG compression using different message lengths at different PSNR values (a) 40 dB and (b) 45 dB
443 Comparative Study To prove the superiority and therobustness advantages of our scheme a comparative studywith state of the art works is indispensable In fact wecompared the robustness of our scheme with six robustwatermarking schemes [26 27 35 37ndash39] In order to carryout a fair comparison the same insertion parameters (PSNRwatermark size) and attacks are used in this comparisonThe PSNRs value of different images and their correspondingvalues in the comparative schemes are illustrated in Figure 12It is worth mentioning that the insertion parameters (119897 119905 119879)were empirically chosen to produce the same PSNR alongsidewith an acceptable visual quality
The robustness results are presented in Tables 2ndash5 interms of either BER or NC depending on how they appearin the reference papers and the superior results are markedin bold number
Based on the simulation results given in Tables 2ndash5 it canbe found that our approach outperforms similar schemes on7194 of the time Equal results are obtained on 2343 ofthe time and our scheme underperforms on 462 of thecases
45 Security The robustness of a watermarking scheme isnot sufficient to judge a scheme as reliable As discussedin the first section robust watermarking is used to secureimages Hence the security is necessary First the watermarkshould be immune against illegal extraction even when theembedding procedure is exposed Second the watermarkdetector should not report the existence of the watermarkwhen none is embedded To evaluate these two risks weextract the watermark from a watermarked image using
Security and Communication Networks 11
Table 2 BER comparison with [27] and [37] using a message length of 128 bits
different keys and extract a watermark from a nonwater-marked image respectively In the two cases the maximumcorrelation obtained has not exceeded the value 059 whichpoints out a negative presence
5 Conclusion
The mean value of DWTrsquos vertical and horizontal subbandscoefficients is highly robust and imperceptible In fact these
two advantages were exploited in the design of a robustand blind watermarking approach The embedding methodconsisted in modulating the BMV in accordance with thequantization technique The use of a single transformationdomain alongside with a simple embedding technique makesfrom the proposed algorithmcomputationally undemandingAt the extraction phase the watermark bits were identifiedby the BMV position with respect to the zero value Unlikesome existing schemes the extraction rules did not require to
Security and Communication Networks 15
be adjusted to the applied attacks to fulfill a high robustnessIn fact the inserted watermark was capable to withstand withno error after JPEG compression up to a quality factor of 30Gaussian filter and other attacks
Data Availability
The data used to support the findings of this study areincluded within the article
Conflicts of Interest
The authors declare that they have no conflicts of interest
References
[1] S Liansheng Z Bei W Zhanmin and T Ailing ldquoAn opticalcolor image watermarking scheme by using compressive sens-ingwith human visual characteristics in gyrator domainrdquoOpticsand Lasers in Engineering vol 92 pp 85ndash93 2017
[2] D Shukla and M Sharma ldquoA Novel Scene-Based Video Water-marking Scheme for Copyright Protectionrdquo Journal of IntelligentSystems vol 27 no 1 pp 47ndash66 2018
[3] A Tiwari and M Sharma ldquoAn efficient vector quantizationbased watermarking method for image integrity authentica-tionrdquo Advances in Intelligent Systems and Computing vol 519pp 215ndash225 2018
[4] S Bravo-Solorio F Calderon C-T Li and A K Nandi ldquoFastfragile watermark embedding and iterative mechanism withhigh self-restoration performancerdquo Digital Signal Processingvol 73 pp 83ndash92 2018
[5] A Ghardallou A Kricha A Sakly and A Mtibaa ldquoAdaptiveblock sized reversible watermarking scheme based on integertransformrdquo in Proceedings of the 17th International Conferenceon Sciences and Techniques of Automatic Control and ComputerEngineering STA 2016 pp 347ndash351 Tunisia December 2016
[6] R G van Schyndel A Z Tirkel and C F Osborne ldquoA digitalwatermarkrdquo in Proceedings of the IEEE International ConferenceImage Processing (ICIP rsquo94) vol 2 pp 86ndash90 Austin Tex USANovember 1994
[7] W-H Lin S-J Horng T-W Kao P Fan C-L Lee and Y PanldquoAn efficient watermarking method based on significant differ-ence of wavelet coefficient quantizationrdquo IEEE Transactions onMultimedia vol 10 no 5 pp 746ndash757 2008
[8] S Bhattacharyya and G Sanyal ldquoA Robust Image Steganogra-phy using DWT Difference Modulation (DWTDM)rdquo Interna-tional Journal of Computer Network and Information Securityvol 4 no 7 pp 27ndash40 2012
[9] H S Malvar and D A Florencio ldquoImproved spread spectruma new modulation technique for robust watermarkingrdquo IEEETransactions on Signal Processing vol 51 no 4 pp 898ndash9052003
[10] B Chen and G W Wornell ldquoProvably robust digital water-markingrdquo in Proceedings of the 1999 Multimedia Systems andApplications II pp 43ndash54 September 1999
[11] Z Kricha A Kricha and A Sakly ldquoAccommodative extractorfor QIM-based watermarking schemesrdquo IET Image Processing2018
[12] I J Cox J Kilian T Leighton and T Shamoon ldquoSecure imper-ceptable yet perceptually salient spread spectrum watermark
formultimediardquo in Proceedings of the 1996 Southcon Conferencepp 192ndash197 June 1996
[13] X Li J Liu J Sun Z Sun H Hu and L Zhang ldquoImprovedspread transform dither modulation-based watermark algo-rithm based on step projectionrdquo in Proceedings of the 2010 IEEEInternational Symposium on Broadband Multimedia Systemsand Broadcasting (BMSB) pp 1ndash5 Shanghai China March2010
[14] H-T Hu and L-Y Hsu ldquoExploring DWT-SVD-DCT featureparameters for robust multiple watermarking against JPEG andJPEG2000 compressionrdquo Computers and Electrical Engineeringvol 41 pp 52ndash63 2015
[15] J Cannons and P Moulin ldquoDesign and statistical analysis of ahash-aided image watermarking systemrdquo IEEE Transactions onImage Processing vol 13 no 10 pp 1393ndash1408 2004
[16] A Valizadeh and Z J Wang ldquoAn improved multiplicativespread spectrum embedding scheme for data hidingrdquo IEEETransactions on Information Forensics and Security vol 7 no4 pp 1127ndash1143 2012
[17] M Kutter D Frederic and F B JordanDigital signature of colorimages using amplitude modulation 1997
[18] Q Su and B Chen ldquoRobust color image watermarking tech-nique in the spatial domainrdquo Soft Computing vol 22 no 1 pp91ndash106 2018
[19] Z Yanling Z Xiaoshi L Na et al ldquoA Digital ImageWatermarkAlgorithm Based onDCCoefficientsQuantizationrdquo in Proceed-ings of the 2006 6th World Congress on Intelligent Control andAutomation pp 9734ndash9738 Dalian China 2006
[20] RMZhaoH LianHWPang andBNHu ldquoAWatermarkingAlgorithm by Modifying AC Coefficies in DCT Domainrdquo inProceedings of the 2008 International Symposium on InformationScience and Engineering (ISISE) pp 159ndash162 Shanghai ChinaDecember 2008
[21] B L Gunjal and R RManthalkar ldquoDiscreteWavelet Transformbased strongly robust watermarking scheme for informationhiding in digital imagesrdquo in Proceedings of the 3rd InternationalConference on Emerging Trends in Engineering and TechnologyICETET 2010 pp 124ndash129 India November 2010
[22] H Shi and F Lv ldquoA Blind Digital Watermark Techniquefor Color Image Based on Integer Wavelet Transformrdquo inProceedings of the 2010 International Conference on BiomedicalEngineering and Computer Science (ICBECS) pp 1ndash4 WuhanChina April 2010
[23] S Ranjbar F Zargari and M Ghanbari ldquoA highly robust two-stage Contourlet-based digital image watermarking methodrdquoSignal Processing Image Communication vol 28 no 10 pp1526ndash1536 2013
[24] Q Su and B Chen ldquoA novel blind color image watermarkingusing upper Hessenbergmatrixrdquo AEU - International Journal ofElectronics and Communications vol 78 pp 64ndash71 2017
[25] F Ourique V Licks R Jordan and F Perez-Gonzalez ldquoAngleQIM a novel watermark embedding scheme robust againstamplitude scaling distortionsrdquo in Proceedings of the IEEE Inter-national Conference on Acoustics Speech and Signal Processing(ICASSP rsquo05) pp 797ndash800 March 2005
[26] H R Fazlali S Samavi N Karimi and S Shirani ldquoAdaptiveblind image watermarking using edge pixel concentrationrdquoMultimedia Tools and Applications vol 76 no 2 pp 3105ndash31202017
[27] X-Y Wang Y-N Liu H Xu A-L Wang and H-Y YangldquoBlind optimum detector for robust image watermarking in
16 Security and Communication Networks
nonsubsampled shearlet Domainrdquo Information Sciences vol372 pp 634ndash654 2016
[28] X Liu C Lin and S Yuan ldquoBlind Dual Watermarking forColor Imagesrsquo Authentication and Copyright Protectionrdquo IEEETransactions on Circuits and Systems for Video Technology vol28 no 5 pp 1047ndash1055 2018
[29] Poonam and S M Arora ldquoA DWT-SVD based Robust DigitalWatermarking for Digital Imagesrdquo Procedia Computer Sciencevol 132 pp 1441ndash1448 2018
[30] S Fazli and M Moeini ldquoA robust image watermarking methodbased on DWT DCT and SVD using a new technique forcorrection of main geometric attacksrdquo Optik - InternationalJournal for Light and ElectronOptics vol 127 no 2 pp 964ndash9722016
[31] K Kuppusamy and K Thamodaran ldquoOptimized Image Water-marking Scheme Based On PSOrdquo Procedia Engineering vol 38pp 493ndash503 2012
[32] V S Rao R S Shekhawat and V K Srivastava ldquoA dwt-dct-svd based digital image watermarking scheme using particleswarm optimizationrdquo in Proceedings of the 2012 IEEE StudentsrsquoConference on Electrical Electronics and Computer Science pp1ndash4 2012
[33] T T Takore P R Kumar and G L Devi ldquoA modified blindimage watermarking scheme based on DWT DCT and SVDdomain using GA to optimize robustnessrdquo in Proceedings ofthe 2016 International Conference on Electrical Electronics andOptimization Techniques ICEEOT 2016 pp 2725ndash2729 IndiaMarch 2016
[34] H Hu S Lin and L Hsu ldquoEffective blind speech watermarkingvia adaptive mean modulation and package synchronization inDWT domainrdquo EURASIP Journal on Audio Speech and MusicProcessing vol 2017 no 1 2017
[35] W-H Lin Y-R Wang S-J Horng T-W Kao and Y Pan ldquoAblind watermarking method using maximum wavelet coeffi-cient quantizationrdquo Expert Systems with Applications vol 36no 9 pp 11509ndash11516 2009
[36] X Li J Liu J Sun X Yang andW Liu ldquoStep-projection-basedspread transform dither modulationrdquo IET Information Securityvol 5 no 3 pp 170ndash180 2011
[37] M Hamghalam S Mirzakuchaki andM A Akhaee ldquoGeomet-ric modelling of the wavelet coefficients for image watermark-ing using optimum detectorrdquo IET Image Processing vol 8 no 3pp 162ndash172 2014
[38] V S Verma R K Jha and A Ojha ldquoDigital watermarkextraction using support vector machine with principal com-ponent analysis based feature reductionrdquo Journal of VisualCommunication and Image Representation vol 31 pp 75ndash852015
[39] C Li Z Zhang Y Wang B Ma and D Huang ldquoDithermodulation of significant amplitude difference for waveletbased robust watermarkingrdquo Neurocomputing vol 166 pp404ndash415 2015
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Navigation and Observation
International Journal of
Hindawi
wwwhindawicom Volume 2018
Advances in
Multimedia
Submit your manuscripts atwwwhindawicom
4 Security and Communication Networks
(a) (b)
Figure 2 Lena image (a) original and (b) after modification of all the1198671198711 and 1198711198671 coefficients
0010203040506070809
1
CDF
|BMV|
0 10 20 30 40 50 60 70 80
(a)
0 2000400 800 1200 1600BMV variation ()
0
01
02
03
04
05
06
07
08
09CD
F
(b)
BMV variation ()
0
01
02
03
04
05
06
07
08
09
CDF
0 10 20 30 40 50 60 70 80 90 100
(c)
0
01
02
03
04
05
06
07
08
09
BMV variation ()
CDF
0 10 20 30 40 50 60 70 80 90 100
(d)
Figure 3 Absolute value of Cumulative histogram of (a) original BMVs (b) difference between the original BMVs and the altered ones ((a)and (b)) (c) difference after JPEG(30) compression and (d) difference after median filtering (3 times 3)
It is worth noting that despite the robustness and imper-ceptibility of the BMVs the final results will also depend onthe insertion and detection method described in the nextsection
3 The Proposed Scheme
This section describes the insertion and extraction processesof a binary watermark 119872 = 1198871 1198872 119887119898 in a grayscale
Security and Communication Networks 5
2timesℎ
2 times w
LHl
HLl
c1 c2 c3 cn
c1 c2 c3 cn
x0
b1 b2 b3 bm
Original Image
Host coefficientsShuffling Chaotic sequence
generation
Block partitioning
Block meanmodulation
Watermark bits
Reshuffling
Watermarked subbands
Watermarked coefficients
BL1 BL2BLlfloorntrfloor
BL1 BL
2 BL
m
BLm+1BLlfloorntrfloor
LHl
HLl
c1c2
c3cn
Figure 4 Block diagram of the watermark embedding scheme
image I with 2 times ℎ height and 2 times 119908 width At this point theembedding parameters are not specified but kept generalizedto point out the flexibility of the scheme At first the originalimage goes through 119897 levels of DWT in order to manifest thedifferent frequencies of the image For the reasons presentedin Section 2 the 119871119867119897 and the 119867119871 119897 subbands are selected forembedding Let 119871119867119897 = 119871119867(119894 119895) and 119867119871 119897 = 119867119871(119894 119895) bethe full set of carrier coefficients where 119894 isin [1 1199082119897minus1] and119895 isin [1 ℎ2119897minus1] The coefficients of the two subbands are thenconcatenated into a single vector 119862 = 119888119896 where 119896 isin [1 119899]and 119899 = ℎ times 11990822119897minus3 Next the carrier vector 119862 is shuffledusing a chaotic sequence generated using the logistic mapand eachwatermark bit is embedded bymodulating themeanvalue of a set of coefficients The complete block diagramof the embedding process is given in Figure 4 The threemain stages involved in the watermarking process (chaoticsequence generation coefficients shuffling and mean blockmodulation) are detailed in the rest of this section
31 Chaotic Sequence Generation Chaotic maps are dynamicsystems defined by a mathematical function ((119909119899+1 = 119891(119909119899))that model a chaotic behavior In practice a variety of chaoticsystems may be utilized (eg Bakerrsquos map circle map duffingmap logistic map etc) The implementation of the latter isone of the simplest thanks to its polynomial mapping degreethough itrsquos high dynamic complexity These features allowedthe logistic map to attract a lot of attention in the informationsecurity disciplines such as cryptography and watermarking
The logistic map may be described by the followingrecurrence relation
119909119899+1 = 119891 (119909119899) = 120582 (119909119899 minus 1199092119899) (1)
The chaotic sequence 119909119899 generated by the function 119891 after119899 iterations is determined by its input parameters 120582 and 1199090where 1199090 is the initial value of the sequence and 120582 is thecontrol parameters In practice 1199090 is choosing between [0 1]and120582 lies between 3569955672 and 4These ranges guaranteea chaotic sequence with a high randomness in the outputwhere the slightest difference in the initial parameters (1199090 and120582) will lead to uncorrelated sequences over time
32 Shuffling In the shuffling phase the elements of the vec-tor 119862 are assigned to new positions according the generatedchaotic sequence The objective of this shuffling is double-fold Firstly it aims to improve the security of the schemethe 119905 coefficients forming each block (119861119871) are randomlyselected therebymaking it difficult for a third party to extractthe watermark or selectively attack the host coefficients todestroy the watermark without the permutation sequenceSecondly the shuffling dispels the correlation between adja-cent coefficients In fact adjacent coefficients tend to beof the same sign leading to MVBs distant from zero Incontrast the MVBs of scattered coefficients are gatheredaround the zero value which will serve us to minimizethe embedding strength and hence the perceptibility of thewatermark
The logistic map defined by (1) is employed to producea pseudo-random key containing the new positions for eachcoefficient
Let 119909119899 = (1199091 1199092 1199093 119909119899) be the chaotic sequencegenerated by means of the logistic map using the secretkey SK = 1199090 120582 where 119899 is the number of iterationsthat is equal to the total number of coefficients in 119871119867119897 and119867119871 119897 The 119909119899 coefficients are then sorted in a ascending
6 Security and Communication Networks
LL
HH
12
Chaotic sequence
4 2 7 1 6 5 3 8
12 09 -03 01 -24 -11 -07 -18
01 09 -07 12 -11 -24 -03 -18
022 025 028 067 074 081 090 0914 2 7 1 6 5 3 8
067 025 090 022 081 074 028 0911 2 3 4 5 6 7 8
Permutation sequence
Sorting09
-03 01
-24 -11
-07 -18C
C
Figure 5 Shuffling of DWT diagonal subbands using a chaotic sequence of eight real numbers
order (1199091199041 1199091199042 1199091199043 119909119904119899) and so the permutation sequence(1199041 1199042 1199043 119904119899) is obtained
The example shown in Figure 5 illustrates a concreteexample for the shuffling operation In this example 119899 is equalto eight the vector 119862 is the concatenation of the diagonalDWT coefficients and 119862 is the shuffled version of 11986233 Block Mean Modulation The first step in the blockmodulation phase is the partitioning of the confused hostcoefficients 119888119894 into independent blocks This operation isrealized by gathering each 119905 adjacent coefficients into a singleblock 119861119871 119894 as given by (2) where 119894 is the index of the blockand is the concatenation symbol The total number ofgenerated blocks 119861119871 should be larger or equal than the binarywatermark message length In other words lfloor119899119905rfloor ge 119898 suchas 119905 isin Nlowast
To embed a robust watermark we consider the quantizationtechnique which modulates the carrier hosts to 119875 nonover-lapping sets where 119875 represents the possible embeddedwords In this work the watermark is a binary messagehence the carrier will be modulated to two different sets(119901 = 2) each associated to a watermark bit The robustnessof the quantization-based insertion method depends on thedistance between the two sets On the other hand wideningthe gap will lead tomediocre results in terms of visual qualityTo work around this issue Chen et al [10] proposed toincrease the element of each set and it is up to the embeddingalgorithm to choose the appropriate quantizer among theavailable ones However as illustrated in Figure 3(a) mostof the BMV are concentrated around the zero where in 80of the BMVs are in the range [minus20 20] Thus we propose tomodulate the blocksrsquo mean values by a positive or negativenumber according to the embedded bit Let 119861119871 119894 be the meanvalue of the block 119894 defined by
119861119871 119894 = 1119905119894times119905sum
119895=119905(119894minus1)+1
119888119895 (3)
The modulation of 119861119871 119894 goes by firstly nullifying the meanvalue by spreading its value equally to all blockrsquos coefficientsthen adding or subtracting a threshold 119879 according to theembedded bit 119887 The equation of modulation is given by (4)where 119888119896 is a coefficient belonging to the block 119894 (119861119871 119894)
1198881015840119896 = 119888119896 minus 119861119871 119894119905 minus 119879 if 119887 = 0119888119896 minus 119861119871 119894119905 + 119879 otherwise (4)
Once all the watermark bits are inserted the watermarkedcoefficients are rearranged into their initial positions usingthe same key SK Finally the inverse DWT is performed toobtain the watermarked image
34 Watermark Extraction In the watermark extractionphase the watermarked image is firstly transformed in thefrequency domain as in the embedding and the same key SKis used to constitute different blocks The decision upon theembedded bit is judged according to the blockrsquos mean valueof the attacked watermarked block 119861119871119894
119887 = 1 if 119861119871119894 gt 00 otherwise (5)
4 Experimental Results
This section is dedicated to evaluate the performance ofthe suggested watermarking framework A variety of 8-bitdepth (119889 = 8) grayscale standard benchmark images witha resolution of 512 times 512 pixels (119908 = ℎ = 256) have beensubjects of assessment The used image database includesLena Baboon Peppers and Tank These images present avariety of image properties (eg mixture of dark and lightarea alongside with highly textures regions (hair) in the Lenaimage) which refutes the idea of tying the performance of thescheme to some image properties
The performance of the suggested method is experimen-tally examined from different aspects Firstly the relation
Security and Communication Networks 7
38 41 45PSNR (dB)
2
4
6
8
10
12
14
16
18
Thre
shol
d va
lue (
T)
123
DWT level
Figure 6 Relation of threshold value (T) and the PSNR at different DWT levels
between the capacity and the imperceptibility is evaluatedusing different insertion parameters Next the evolution ofcomplexity according to the host image size is monitoredAfterwards the robustness is validated in two ways initiallythe robustness against different levels of attacks is examinedand then a comparative study with similar works is presentedFinally the security of the scheme is verified
41 Imperceptibility The imperceptibility feature reflects thefidelity of the watermarked image toward the original oneSeveral metrics can be used to quantify the noticeable differ-ence between two images involving subjective and objectivemetrics In this work we employed the Peek Signal to NoiseRatio (PSNR) and the Structural Similarity Index Measure(SSIM) given by (6) and (7) respectively
PSNR (I Ilowast) = 10 log10( (2119889 minus 1)2 times 119908 times ℎsum119908119894=1sumℎ119894=1 (I119894119895 minus Ilowast119894119895)2) (6)
where 120579I and 120583I are respectively the mean intensity and thestandard deviation of the original image I and 120590119868119868lowast is thesample cross-correlation between I and Ilowast The constants 1198781and 1198782 are defined as 1198781 = (1198961(2119889minus1))2 and 1198782 = (1198962(2119889minus1))2where 1198961 1198962 ≪ 142 Capacity The maximum capacity of the proposedscheme depends on three variables the image dimension (119908
and ℎ) the wavelet level (119897) and the block length (119905) Hencewe can express the maximum capacity using
Max capacity = lfloor 119908 times ℎ119905 times 22119897minus3rfloor (8)
For a better understanding of the relation between the capac-ity and imperceptibility we embedded a variable message size(119898 isin 256 512 1024) at different levels of decomposition(119897 isin 1 2 3) into three different images namely Lena Coupleand Baboon For each watermarking operation the blocksrsquosize 119905 is maximized so that all coefficients are involved inthe watermarking process (eg 119897 = 2 and 119905 = 32 for119898 = 1024) The threshold 119879 is manually chosen to reach aPSNR value of 38 41 and 45 dBThe parameters of insertion(119879 119905 119897 119898) the objective measurements (PSNR and SSIM)and the watermarked images are all listed in Table 1
The first observation that we draw from Table 1 is thatfor a given PSNR the threshold 119879 is converging towardthe same value regardless the host image or the watermarksize This amounts to saying that we can anticipate the119879 value according to the level of transformation and thetargeted PSNR Figure 6 illustrates the relation between theseparameters where119879 is obtained by averaging all the values fora specific level and PSNR Secondly it is shown from Table 1that at 45 dB the distortion due to the watermark insertionis invisible whatever the size of the message and the levelof decomposition However for watermarked images with aPSNR of 38 and 41 dB the imperceptibility is conditionedupon the level 119897 the higher 119897 is the lower the quality is goingto be
43 Complexity The computational cost of the watermarkembedding and extraction is an important aspect that may
8 Security and Communication Networks
Table 1 Quality of watermarked images using different insertion parameters (119897 119879119898)Message size Transformation
level Watermarked Lena images Watermarked Coupleimages
Figure 7 Computational time of (a) overall insertion and detection and (b) details of the embedding stage
restrict or expand the use of a given watermarking schemeIn the literature most of schemes are evaluated in a standardimage with a size of 512 times 512 pixels and there is noperspective of the computational cost trend when the imagegoes bigger which is more realistic Hence we propose tovary the size of the image beginning with 64 times 64 pixelsand ending with 4096 times 4096 pixels by doubling the heightand width after each insertion For the sake of simplicity thewatermark was embedded after one level of DWT (119897 = 1)with a fixed block size (119905 = 4) and a maximum watermarksize as defined by (8)
Figure 7(a) depicts the execution time evolution ofthe overall embedding and detection alongside with thevariation of the watermark size (red) as a function of theimage size The details of the embedding phase are shownin Figure 7(b) The most important remark to be made fromFigure 7 is that the computational time evolves broadly inline with the watermark size (red curve) to find out thecomplexity of our scheme in comparison with other state-ofthe-art methods
44 Robustness In the following we investigate the robust-ness of the suggested watermarking approach under severaltypes of attacks The robustness is quantified using the BitError Rate (BER) and the Normalized Correlation (NC)metrics
441 Evaluated Attacks Since the proposed scheme isdesigned to be robust the embedded message should lastor at list a significant part of it after image processingoperations Fromwatermark perspective all image processingoperations whatever performed on goodwill or not is consid-ered as attacks The involved attacks in the evaluation tests areclassified into three groups
(i) Compression attacks usually images are not storedin a raw format but after a compression operationthat aims to reduce the required storage space JPEGalgorithm is one of the most used compression
algorithms However JPEG is a lossy algorithm thatcauses some quality degradation inversely commen-surate with the compression quality factor Hence wetested the robustness of our algorithm under differentJPEG quality factors ranging from 10 to 90
(ii) Image processing attacks used to imitate a noisecontaminating the image during transmission such assalt and peppers or to improve the imagersquos qualityby eliminating a noise or enhancing some propertiesThe used attacks are listed as follows scaling medianfiltering sharpening Gaussian filtering average fil-tering Gaussian noise salt and pepper noise contam-ination histogram equalization and additive whiteGaussian noise (AWGN)
(iii) Geometric attacks used to correct geometric distor-tions such camera orientation (rotation) or imagesrsquoalignment (cropping) The used attacks are listed asfollows image cropping and resizing and rotation
Since the DWT coefficients are not geometrically invari-ant a manual synchronization of the imagersquos dimension andorientation was realized prior to the extraction
442 Performance Evaluation Here we test the robustnessof our scheme on three different images Boat Peppersand Barbara Each time we varied the embedded messagebetween 128 bits and 2048 bits The embedding threshold 119879is manually tuned to provide a watermarked image with aPSNR value of 40 dB and 45 dB The robustness is expressedin terms of BER () after mean filter (Figure 8) JPEGcompression (Figure 9) median filter (Figure 10) and gammacorrection Poisson noise histogram equalization and imageunsharpening (Figure 11) Figures 8ndash11 show that the robust-ness performance slightly differs between images For a smallmessage length (between 128 and 256 bits) and a PSNR valueof 40 dB the BER does not exceed 20 for most of attacksHowever the larger the message size is or the stronger theattack is the higher the BER will be
10 Security and Communication Networks
128 256 512 1024 2048Watermark size
0
10
20
30
40
50BE
R (
)
Mean filter (3x3)Mean filter (5x5)Mean filter (7x7)Mean filter (9x9)BarbaraPeppersBoat
(a)
Mean filter (3x3)Mean filter (5x5)Mean filter (7x7)Mean filter (9x9)BarbaraPeppersBoat
128 256 512 1024 2048Watermark size
0
10
20
30
40
50
BER
()
(b)
Figure 8 BER() after mean filter attacks using different message lengths at different PSNR values (a) 40 dB and (b) 45 dB
Figure 9 BER() after JPEG compression using different message lengths at different PSNR values (a) 40 dB and (b) 45 dB
443 Comparative Study To prove the superiority and therobustness advantages of our scheme a comparative studywith state of the art works is indispensable In fact wecompared the robustness of our scheme with six robustwatermarking schemes [26 27 35 37ndash39] In order to carryout a fair comparison the same insertion parameters (PSNRwatermark size) and attacks are used in this comparisonThe PSNRs value of different images and their correspondingvalues in the comparative schemes are illustrated in Figure 12It is worth mentioning that the insertion parameters (119897 119905 119879)were empirically chosen to produce the same PSNR alongsidewith an acceptable visual quality
The robustness results are presented in Tables 2ndash5 interms of either BER or NC depending on how they appearin the reference papers and the superior results are markedin bold number
Based on the simulation results given in Tables 2ndash5 it canbe found that our approach outperforms similar schemes on7194 of the time Equal results are obtained on 2343 ofthe time and our scheme underperforms on 462 of thecases
45 Security The robustness of a watermarking scheme isnot sufficient to judge a scheme as reliable As discussedin the first section robust watermarking is used to secureimages Hence the security is necessary First the watermarkshould be immune against illegal extraction even when theembedding procedure is exposed Second the watermarkdetector should not report the existence of the watermarkwhen none is embedded To evaluate these two risks weextract the watermark from a watermarked image using
Security and Communication Networks 11
Table 2 BER comparison with [27] and [37] using a message length of 128 bits
different keys and extract a watermark from a nonwater-marked image respectively In the two cases the maximumcorrelation obtained has not exceeded the value 059 whichpoints out a negative presence
5 Conclusion
The mean value of DWTrsquos vertical and horizontal subbandscoefficients is highly robust and imperceptible In fact these
two advantages were exploited in the design of a robustand blind watermarking approach The embedding methodconsisted in modulating the BMV in accordance with thequantization technique The use of a single transformationdomain alongside with a simple embedding technique makesfrom the proposed algorithmcomputationally undemandingAt the extraction phase the watermark bits were identifiedby the BMV position with respect to the zero value Unlikesome existing schemes the extraction rules did not require to
Security and Communication Networks 15
be adjusted to the applied attacks to fulfill a high robustnessIn fact the inserted watermark was capable to withstand withno error after JPEG compression up to a quality factor of 30Gaussian filter and other attacks
Data Availability
The data used to support the findings of this study areincluded within the article
Conflicts of Interest
The authors declare that they have no conflicts of interest
References
[1] S Liansheng Z Bei W Zhanmin and T Ailing ldquoAn opticalcolor image watermarking scheme by using compressive sens-ingwith human visual characteristics in gyrator domainrdquoOpticsand Lasers in Engineering vol 92 pp 85ndash93 2017
[2] D Shukla and M Sharma ldquoA Novel Scene-Based Video Water-marking Scheme for Copyright Protectionrdquo Journal of IntelligentSystems vol 27 no 1 pp 47ndash66 2018
[3] A Tiwari and M Sharma ldquoAn efficient vector quantizationbased watermarking method for image integrity authentica-tionrdquo Advances in Intelligent Systems and Computing vol 519pp 215ndash225 2018
[4] S Bravo-Solorio F Calderon C-T Li and A K Nandi ldquoFastfragile watermark embedding and iterative mechanism withhigh self-restoration performancerdquo Digital Signal Processingvol 73 pp 83ndash92 2018
[5] A Ghardallou A Kricha A Sakly and A Mtibaa ldquoAdaptiveblock sized reversible watermarking scheme based on integertransformrdquo in Proceedings of the 17th International Conferenceon Sciences and Techniques of Automatic Control and ComputerEngineering STA 2016 pp 347ndash351 Tunisia December 2016
[6] R G van Schyndel A Z Tirkel and C F Osborne ldquoA digitalwatermarkrdquo in Proceedings of the IEEE International ConferenceImage Processing (ICIP rsquo94) vol 2 pp 86ndash90 Austin Tex USANovember 1994
[7] W-H Lin S-J Horng T-W Kao P Fan C-L Lee and Y PanldquoAn efficient watermarking method based on significant differ-ence of wavelet coefficient quantizationrdquo IEEE Transactions onMultimedia vol 10 no 5 pp 746ndash757 2008
[8] S Bhattacharyya and G Sanyal ldquoA Robust Image Steganogra-phy using DWT Difference Modulation (DWTDM)rdquo Interna-tional Journal of Computer Network and Information Securityvol 4 no 7 pp 27ndash40 2012
[9] H S Malvar and D A Florencio ldquoImproved spread spectruma new modulation technique for robust watermarkingrdquo IEEETransactions on Signal Processing vol 51 no 4 pp 898ndash9052003
[10] B Chen and G W Wornell ldquoProvably robust digital water-markingrdquo in Proceedings of the 1999 Multimedia Systems andApplications II pp 43ndash54 September 1999
[11] Z Kricha A Kricha and A Sakly ldquoAccommodative extractorfor QIM-based watermarking schemesrdquo IET Image Processing2018
[12] I J Cox J Kilian T Leighton and T Shamoon ldquoSecure imper-ceptable yet perceptually salient spread spectrum watermark
formultimediardquo in Proceedings of the 1996 Southcon Conferencepp 192ndash197 June 1996
[13] X Li J Liu J Sun Z Sun H Hu and L Zhang ldquoImprovedspread transform dither modulation-based watermark algo-rithm based on step projectionrdquo in Proceedings of the 2010 IEEEInternational Symposium on Broadband Multimedia Systemsand Broadcasting (BMSB) pp 1ndash5 Shanghai China March2010
[14] H-T Hu and L-Y Hsu ldquoExploring DWT-SVD-DCT featureparameters for robust multiple watermarking against JPEG andJPEG2000 compressionrdquo Computers and Electrical Engineeringvol 41 pp 52ndash63 2015
[15] J Cannons and P Moulin ldquoDesign and statistical analysis of ahash-aided image watermarking systemrdquo IEEE Transactions onImage Processing vol 13 no 10 pp 1393ndash1408 2004
[16] A Valizadeh and Z J Wang ldquoAn improved multiplicativespread spectrum embedding scheme for data hidingrdquo IEEETransactions on Information Forensics and Security vol 7 no4 pp 1127ndash1143 2012
[17] M Kutter D Frederic and F B JordanDigital signature of colorimages using amplitude modulation 1997
[18] Q Su and B Chen ldquoRobust color image watermarking tech-nique in the spatial domainrdquo Soft Computing vol 22 no 1 pp91ndash106 2018
[19] Z Yanling Z Xiaoshi L Na et al ldquoA Digital ImageWatermarkAlgorithm Based onDCCoefficientsQuantizationrdquo in Proceed-ings of the 2006 6th World Congress on Intelligent Control andAutomation pp 9734ndash9738 Dalian China 2006
[20] RMZhaoH LianHWPang andBNHu ldquoAWatermarkingAlgorithm by Modifying AC Coefficies in DCT Domainrdquo inProceedings of the 2008 International Symposium on InformationScience and Engineering (ISISE) pp 159ndash162 Shanghai ChinaDecember 2008
[21] B L Gunjal and R RManthalkar ldquoDiscreteWavelet Transformbased strongly robust watermarking scheme for informationhiding in digital imagesrdquo in Proceedings of the 3rd InternationalConference on Emerging Trends in Engineering and TechnologyICETET 2010 pp 124ndash129 India November 2010
[22] H Shi and F Lv ldquoA Blind Digital Watermark Techniquefor Color Image Based on Integer Wavelet Transformrdquo inProceedings of the 2010 International Conference on BiomedicalEngineering and Computer Science (ICBECS) pp 1ndash4 WuhanChina April 2010
[23] S Ranjbar F Zargari and M Ghanbari ldquoA highly robust two-stage Contourlet-based digital image watermarking methodrdquoSignal Processing Image Communication vol 28 no 10 pp1526ndash1536 2013
[24] Q Su and B Chen ldquoA novel blind color image watermarkingusing upper Hessenbergmatrixrdquo AEU - International Journal ofElectronics and Communications vol 78 pp 64ndash71 2017
[25] F Ourique V Licks R Jordan and F Perez-Gonzalez ldquoAngleQIM a novel watermark embedding scheme robust againstamplitude scaling distortionsrdquo in Proceedings of the IEEE Inter-national Conference on Acoustics Speech and Signal Processing(ICASSP rsquo05) pp 797ndash800 March 2005
[26] H R Fazlali S Samavi N Karimi and S Shirani ldquoAdaptiveblind image watermarking using edge pixel concentrationrdquoMultimedia Tools and Applications vol 76 no 2 pp 3105ndash31202017
[27] X-Y Wang Y-N Liu H Xu A-L Wang and H-Y YangldquoBlind optimum detector for robust image watermarking in
16 Security and Communication Networks
nonsubsampled shearlet Domainrdquo Information Sciences vol372 pp 634ndash654 2016
[28] X Liu C Lin and S Yuan ldquoBlind Dual Watermarking forColor Imagesrsquo Authentication and Copyright Protectionrdquo IEEETransactions on Circuits and Systems for Video Technology vol28 no 5 pp 1047ndash1055 2018
[29] Poonam and S M Arora ldquoA DWT-SVD based Robust DigitalWatermarking for Digital Imagesrdquo Procedia Computer Sciencevol 132 pp 1441ndash1448 2018
[30] S Fazli and M Moeini ldquoA robust image watermarking methodbased on DWT DCT and SVD using a new technique forcorrection of main geometric attacksrdquo Optik - InternationalJournal for Light and ElectronOptics vol 127 no 2 pp 964ndash9722016
[31] K Kuppusamy and K Thamodaran ldquoOptimized Image Water-marking Scheme Based On PSOrdquo Procedia Engineering vol 38pp 493ndash503 2012
[32] V S Rao R S Shekhawat and V K Srivastava ldquoA dwt-dct-svd based digital image watermarking scheme using particleswarm optimizationrdquo in Proceedings of the 2012 IEEE StudentsrsquoConference on Electrical Electronics and Computer Science pp1ndash4 2012
[33] T T Takore P R Kumar and G L Devi ldquoA modified blindimage watermarking scheme based on DWT DCT and SVDdomain using GA to optimize robustnessrdquo in Proceedings ofthe 2016 International Conference on Electrical Electronics andOptimization Techniques ICEEOT 2016 pp 2725ndash2729 IndiaMarch 2016
[34] H Hu S Lin and L Hsu ldquoEffective blind speech watermarkingvia adaptive mean modulation and package synchronization inDWT domainrdquo EURASIP Journal on Audio Speech and MusicProcessing vol 2017 no 1 2017
[35] W-H Lin Y-R Wang S-J Horng T-W Kao and Y Pan ldquoAblind watermarking method using maximum wavelet coeffi-cient quantizationrdquo Expert Systems with Applications vol 36no 9 pp 11509ndash11516 2009
[36] X Li J Liu J Sun X Yang andW Liu ldquoStep-projection-basedspread transform dither modulationrdquo IET Information Securityvol 5 no 3 pp 170ndash180 2011
[37] M Hamghalam S Mirzakuchaki andM A Akhaee ldquoGeomet-ric modelling of the wavelet coefficients for image watermark-ing using optimum detectorrdquo IET Image Processing vol 8 no 3pp 162ndash172 2014
[38] V S Verma R K Jha and A Ojha ldquoDigital watermarkextraction using support vector machine with principal com-ponent analysis based feature reductionrdquo Journal of VisualCommunication and Image Representation vol 31 pp 75ndash852015
[39] C Li Z Zhang Y Wang B Ma and D Huang ldquoDithermodulation of significant amplitude difference for waveletbased robust watermarkingrdquo Neurocomputing vol 166 pp404ndash415 2015
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Navigation and Observation
International Journal of
Hindawi
wwwhindawicom Volume 2018
Advances in
Multimedia
Submit your manuscripts atwwwhindawicom
Security and Communication Networks 5
2timesℎ
2 times w
LHl
HLl
c1 c2 c3 cn
c1 c2 c3 cn
x0
b1 b2 b3 bm
Original Image
Host coefficientsShuffling Chaotic sequence
generation
Block partitioning
Block meanmodulation
Watermark bits
Reshuffling
Watermarked subbands
Watermarked coefficients
BL1 BL2BLlfloorntrfloor
BL1 BL
2 BL
m
BLm+1BLlfloorntrfloor
LHl
HLl
c1c2
c3cn
Figure 4 Block diagram of the watermark embedding scheme
image I with 2 times ℎ height and 2 times 119908 width At this point theembedding parameters are not specified but kept generalizedto point out the flexibility of the scheme At first the originalimage goes through 119897 levels of DWT in order to manifest thedifferent frequencies of the image For the reasons presentedin Section 2 the 119871119867119897 and the 119867119871 119897 subbands are selected forembedding Let 119871119867119897 = 119871119867(119894 119895) and 119867119871 119897 = 119867119871(119894 119895) bethe full set of carrier coefficients where 119894 isin [1 1199082119897minus1] and119895 isin [1 ℎ2119897minus1] The coefficients of the two subbands are thenconcatenated into a single vector 119862 = 119888119896 where 119896 isin [1 119899]and 119899 = ℎ times 11990822119897minus3 Next the carrier vector 119862 is shuffledusing a chaotic sequence generated using the logistic mapand eachwatermark bit is embedded bymodulating themeanvalue of a set of coefficients The complete block diagramof the embedding process is given in Figure 4 The threemain stages involved in the watermarking process (chaoticsequence generation coefficients shuffling and mean blockmodulation) are detailed in the rest of this section
31 Chaotic Sequence Generation Chaotic maps are dynamicsystems defined by a mathematical function ((119909119899+1 = 119891(119909119899))that model a chaotic behavior In practice a variety of chaoticsystems may be utilized (eg Bakerrsquos map circle map duffingmap logistic map etc) The implementation of the latter isone of the simplest thanks to its polynomial mapping degreethough itrsquos high dynamic complexity These features allowedthe logistic map to attract a lot of attention in the informationsecurity disciplines such as cryptography and watermarking
The logistic map may be described by the followingrecurrence relation
119909119899+1 = 119891 (119909119899) = 120582 (119909119899 minus 1199092119899) (1)
The chaotic sequence 119909119899 generated by the function 119891 after119899 iterations is determined by its input parameters 120582 and 1199090where 1199090 is the initial value of the sequence and 120582 is thecontrol parameters In practice 1199090 is choosing between [0 1]and120582 lies between 3569955672 and 4These ranges guaranteea chaotic sequence with a high randomness in the outputwhere the slightest difference in the initial parameters (1199090 and120582) will lead to uncorrelated sequences over time
32 Shuffling In the shuffling phase the elements of the vec-tor 119862 are assigned to new positions according the generatedchaotic sequence The objective of this shuffling is double-fold Firstly it aims to improve the security of the schemethe 119905 coefficients forming each block (119861119871) are randomlyselected therebymaking it difficult for a third party to extractthe watermark or selectively attack the host coefficients todestroy the watermark without the permutation sequenceSecondly the shuffling dispels the correlation between adja-cent coefficients In fact adjacent coefficients tend to beof the same sign leading to MVBs distant from zero Incontrast the MVBs of scattered coefficients are gatheredaround the zero value which will serve us to minimizethe embedding strength and hence the perceptibility of thewatermark
The logistic map defined by (1) is employed to producea pseudo-random key containing the new positions for eachcoefficient
Let 119909119899 = (1199091 1199092 1199093 119909119899) be the chaotic sequencegenerated by means of the logistic map using the secretkey SK = 1199090 120582 where 119899 is the number of iterationsthat is equal to the total number of coefficients in 119871119867119897 and119867119871 119897 The 119909119899 coefficients are then sorted in a ascending
6 Security and Communication Networks
LL
HH
12
Chaotic sequence
4 2 7 1 6 5 3 8
12 09 -03 01 -24 -11 -07 -18
01 09 -07 12 -11 -24 -03 -18
022 025 028 067 074 081 090 0914 2 7 1 6 5 3 8
067 025 090 022 081 074 028 0911 2 3 4 5 6 7 8
Permutation sequence
Sorting09
-03 01
-24 -11
-07 -18C
C
Figure 5 Shuffling of DWT diagonal subbands using a chaotic sequence of eight real numbers
order (1199091199041 1199091199042 1199091199043 119909119904119899) and so the permutation sequence(1199041 1199042 1199043 119904119899) is obtained
The example shown in Figure 5 illustrates a concreteexample for the shuffling operation In this example 119899 is equalto eight the vector 119862 is the concatenation of the diagonalDWT coefficients and 119862 is the shuffled version of 11986233 Block Mean Modulation The first step in the blockmodulation phase is the partitioning of the confused hostcoefficients 119888119894 into independent blocks This operation isrealized by gathering each 119905 adjacent coefficients into a singleblock 119861119871 119894 as given by (2) where 119894 is the index of the blockand is the concatenation symbol The total number ofgenerated blocks 119861119871 should be larger or equal than the binarywatermark message length In other words lfloor119899119905rfloor ge 119898 suchas 119905 isin Nlowast
To embed a robust watermark we consider the quantizationtechnique which modulates the carrier hosts to 119875 nonover-lapping sets where 119875 represents the possible embeddedwords In this work the watermark is a binary messagehence the carrier will be modulated to two different sets(119901 = 2) each associated to a watermark bit The robustnessof the quantization-based insertion method depends on thedistance between the two sets On the other hand wideningthe gap will lead tomediocre results in terms of visual qualityTo work around this issue Chen et al [10] proposed toincrease the element of each set and it is up to the embeddingalgorithm to choose the appropriate quantizer among theavailable ones However as illustrated in Figure 3(a) mostof the BMV are concentrated around the zero where in 80of the BMVs are in the range [minus20 20] Thus we propose tomodulate the blocksrsquo mean values by a positive or negativenumber according to the embedded bit Let 119861119871 119894 be the meanvalue of the block 119894 defined by
119861119871 119894 = 1119905119894times119905sum
119895=119905(119894minus1)+1
119888119895 (3)
The modulation of 119861119871 119894 goes by firstly nullifying the meanvalue by spreading its value equally to all blockrsquos coefficientsthen adding or subtracting a threshold 119879 according to theembedded bit 119887 The equation of modulation is given by (4)where 119888119896 is a coefficient belonging to the block 119894 (119861119871 119894)
1198881015840119896 = 119888119896 minus 119861119871 119894119905 minus 119879 if 119887 = 0119888119896 minus 119861119871 119894119905 + 119879 otherwise (4)
Once all the watermark bits are inserted the watermarkedcoefficients are rearranged into their initial positions usingthe same key SK Finally the inverse DWT is performed toobtain the watermarked image
34 Watermark Extraction In the watermark extractionphase the watermarked image is firstly transformed in thefrequency domain as in the embedding and the same key SKis used to constitute different blocks The decision upon theembedded bit is judged according to the blockrsquos mean valueof the attacked watermarked block 119861119871119894
119887 = 1 if 119861119871119894 gt 00 otherwise (5)
4 Experimental Results
This section is dedicated to evaluate the performance ofthe suggested watermarking framework A variety of 8-bitdepth (119889 = 8) grayscale standard benchmark images witha resolution of 512 times 512 pixels (119908 = ℎ = 256) have beensubjects of assessment The used image database includesLena Baboon Peppers and Tank These images present avariety of image properties (eg mixture of dark and lightarea alongside with highly textures regions (hair) in the Lenaimage) which refutes the idea of tying the performance of thescheme to some image properties
The performance of the suggested method is experimen-tally examined from different aspects Firstly the relation
Security and Communication Networks 7
38 41 45PSNR (dB)
2
4
6
8
10
12
14
16
18
Thre
shol
d va
lue (
T)
123
DWT level
Figure 6 Relation of threshold value (T) and the PSNR at different DWT levels
between the capacity and the imperceptibility is evaluatedusing different insertion parameters Next the evolution ofcomplexity according to the host image size is monitoredAfterwards the robustness is validated in two ways initiallythe robustness against different levels of attacks is examinedand then a comparative study with similar works is presentedFinally the security of the scheme is verified
41 Imperceptibility The imperceptibility feature reflects thefidelity of the watermarked image toward the original oneSeveral metrics can be used to quantify the noticeable differ-ence between two images involving subjective and objectivemetrics In this work we employed the Peek Signal to NoiseRatio (PSNR) and the Structural Similarity Index Measure(SSIM) given by (6) and (7) respectively
PSNR (I Ilowast) = 10 log10( (2119889 minus 1)2 times 119908 times ℎsum119908119894=1sumℎ119894=1 (I119894119895 minus Ilowast119894119895)2) (6)
where 120579I and 120583I are respectively the mean intensity and thestandard deviation of the original image I and 120590119868119868lowast is thesample cross-correlation between I and Ilowast The constants 1198781and 1198782 are defined as 1198781 = (1198961(2119889minus1))2 and 1198782 = (1198962(2119889minus1))2where 1198961 1198962 ≪ 142 Capacity The maximum capacity of the proposedscheme depends on three variables the image dimension (119908
and ℎ) the wavelet level (119897) and the block length (119905) Hencewe can express the maximum capacity using
Max capacity = lfloor 119908 times ℎ119905 times 22119897minus3rfloor (8)
For a better understanding of the relation between the capac-ity and imperceptibility we embedded a variable message size(119898 isin 256 512 1024) at different levels of decomposition(119897 isin 1 2 3) into three different images namely Lena Coupleand Baboon For each watermarking operation the blocksrsquosize 119905 is maximized so that all coefficients are involved inthe watermarking process (eg 119897 = 2 and 119905 = 32 for119898 = 1024) The threshold 119879 is manually chosen to reach aPSNR value of 38 41 and 45 dBThe parameters of insertion(119879 119905 119897 119898) the objective measurements (PSNR and SSIM)and the watermarked images are all listed in Table 1
The first observation that we draw from Table 1 is thatfor a given PSNR the threshold 119879 is converging towardthe same value regardless the host image or the watermarksize This amounts to saying that we can anticipate the119879 value according to the level of transformation and thetargeted PSNR Figure 6 illustrates the relation between theseparameters where119879 is obtained by averaging all the values fora specific level and PSNR Secondly it is shown from Table 1that at 45 dB the distortion due to the watermark insertionis invisible whatever the size of the message and the levelof decomposition However for watermarked images with aPSNR of 38 and 41 dB the imperceptibility is conditionedupon the level 119897 the higher 119897 is the lower the quality is goingto be
43 Complexity The computational cost of the watermarkembedding and extraction is an important aspect that may
8 Security and Communication Networks
Table 1 Quality of watermarked images using different insertion parameters (119897 119879119898)Message size Transformation
level Watermarked Lena images Watermarked Coupleimages
Figure 7 Computational time of (a) overall insertion and detection and (b) details of the embedding stage
restrict or expand the use of a given watermarking schemeIn the literature most of schemes are evaluated in a standardimage with a size of 512 times 512 pixels and there is noperspective of the computational cost trend when the imagegoes bigger which is more realistic Hence we propose tovary the size of the image beginning with 64 times 64 pixelsand ending with 4096 times 4096 pixels by doubling the heightand width after each insertion For the sake of simplicity thewatermark was embedded after one level of DWT (119897 = 1)with a fixed block size (119905 = 4) and a maximum watermarksize as defined by (8)
Figure 7(a) depicts the execution time evolution ofthe overall embedding and detection alongside with thevariation of the watermark size (red) as a function of theimage size The details of the embedding phase are shownin Figure 7(b) The most important remark to be made fromFigure 7 is that the computational time evolves broadly inline with the watermark size (red curve) to find out thecomplexity of our scheme in comparison with other state-ofthe-art methods
44 Robustness In the following we investigate the robust-ness of the suggested watermarking approach under severaltypes of attacks The robustness is quantified using the BitError Rate (BER) and the Normalized Correlation (NC)metrics
441 Evaluated Attacks Since the proposed scheme isdesigned to be robust the embedded message should lastor at list a significant part of it after image processingoperations Fromwatermark perspective all image processingoperations whatever performed on goodwill or not is consid-ered as attacks The involved attacks in the evaluation tests areclassified into three groups
(i) Compression attacks usually images are not storedin a raw format but after a compression operationthat aims to reduce the required storage space JPEGalgorithm is one of the most used compression
algorithms However JPEG is a lossy algorithm thatcauses some quality degradation inversely commen-surate with the compression quality factor Hence wetested the robustness of our algorithm under differentJPEG quality factors ranging from 10 to 90
(ii) Image processing attacks used to imitate a noisecontaminating the image during transmission such assalt and peppers or to improve the imagersquos qualityby eliminating a noise or enhancing some propertiesThe used attacks are listed as follows scaling medianfiltering sharpening Gaussian filtering average fil-tering Gaussian noise salt and pepper noise contam-ination histogram equalization and additive whiteGaussian noise (AWGN)
(iii) Geometric attacks used to correct geometric distor-tions such camera orientation (rotation) or imagesrsquoalignment (cropping) The used attacks are listed asfollows image cropping and resizing and rotation
Since the DWT coefficients are not geometrically invari-ant a manual synchronization of the imagersquos dimension andorientation was realized prior to the extraction
442 Performance Evaluation Here we test the robustnessof our scheme on three different images Boat Peppersand Barbara Each time we varied the embedded messagebetween 128 bits and 2048 bits The embedding threshold 119879is manually tuned to provide a watermarked image with aPSNR value of 40 dB and 45 dB The robustness is expressedin terms of BER () after mean filter (Figure 8) JPEGcompression (Figure 9) median filter (Figure 10) and gammacorrection Poisson noise histogram equalization and imageunsharpening (Figure 11) Figures 8ndash11 show that the robust-ness performance slightly differs between images For a smallmessage length (between 128 and 256 bits) and a PSNR valueof 40 dB the BER does not exceed 20 for most of attacksHowever the larger the message size is or the stronger theattack is the higher the BER will be
10 Security and Communication Networks
128 256 512 1024 2048Watermark size
0
10
20
30
40
50BE
R (
)
Mean filter (3x3)Mean filter (5x5)Mean filter (7x7)Mean filter (9x9)BarbaraPeppersBoat
(a)
Mean filter (3x3)Mean filter (5x5)Mean filter (7x7)Mean filter (9x9)BarbaraPeppersBoat
128 256 512 1024 2048Watermark size
0
10
20
30
40
50
BER
()
(b)
Figure 8 BER() after mean filter attacks using different message lengths at different PSNR values (a) 40 dB and (b) 45 dB
Figure 9 BER() after JPEG compression using different message lengths at different PSNR values (a) 40 dB and (b) 45 dB
443 Comparative Study To prove the superiority and therobustness advantages of our scheme a comparative studywith state of the art works is indispensable In fact wecompared the robustness of our scheme with six robustwatermarking schemes [26 27 35 37ndash39] In order to carryout a fair comparison the same insertion parameters (PSNRwatermark size) and attacks are used in this comparisonThe PSNRs value of different images and their correspondingvalues in the comparative schemes are illustrated in Figure 12It is worth mentioning that the insertion parameters (119897 119905 119879)were empirically chosen to produce the same PSNR alongsidewith an acceptable visual quality
The robustness results are presented in Tables 2ndash5 interms of either BER or NC depending on how they appearin the reference papers and the superior results are markedin bold number
Based on the simulation results given in Tables 2ndash5 it canbe found that our approach outperforms similar schemes on7194 of the time Equal results are obtained on 2343 ofthe time and our scheme underperforms on 462 of thecases
45 Security The robustness of a watermarking scheme isnot sufficient to judge a scheme as reliable As discussedin the first section robust watermarking is used to secureimages Hence the security is necessary First the watermarkshould be immune against illegal extraction even when theembedding procedure is exposed Second the watermarkdetector should not report the existence of the watermarkwhen none is embedded To evaluate these two risks weextract the watermark from a watermarked image using
Security and Communication Networks 11
Table 2 BER comparison with [27] and [37] using a message length of 128 bits
different keys and extract a watermark from a nonwater-marked image respectively In the two cases the maximumcorrelation obtained has not exceeded the value 059 whichpoints out a negative presence
5 Conclusion
The mean value of DWTrsquos vertical and horizontal subbandscoefficients is highly robust and imperceptible In fact these
two advantages were exploited in the design of a robustand blind watermarking approach The embedding methodconsisted in modulating the BMV in accordance with thequantization technique The use of a single transformationdomain alongside with a simple embedding technique makesfrom the proposed algorithmcomputationally undemandingAt the extraction phase the watermark bits were identifiedby the BMV position with respect to the zero value Unlikesome existing schemes the extraction rules did not require to
Security and Communication Networks 15
be adjusted to the applied attacks to fulfill a high robustnessIn fact the inserted watermark was capable to withstand withno error after JPEG compression up to a quality factor of 30Gaussian filter and other attacks
Data Availability
The data used to support the findings of this study areincluded within the article
Conflicts of Interest
The authors declare that they have no conflicts of interest
References
[1] S Liansheng Z Bei W Zhanmin and T Ailing ldquoAn opticalcolor image watermarking scheme by using compressive sens-ingwith human visual characteristics in gyrator domainrdquoOpticsand Lasers in Engineering vol 92 pp 85ndash93 2017
[2] D Shukla and M Sharma ldquoA Novel Scene-Based Video Water-marking Scheme for Copyright Protectionrdquo Journal of IntelligentSystems vol 27 no 1 pp 47ndash66 2018
[3] A Tiwari and M Sharma ldquoAn efficient vector quantizationbased watermarking method for image integrity authentica-tionrdquo Advances in Intelligent Systems and Computing vol 519pp 215ndash225 2018
[4] S Bravo-Solorio F Calderon C-T Li and A K Nandi ldquoFastfragile watermark embedding and iterative mechanism withhigh self-restoration performancerdquo Digital Signal Processingvol 73 pp 83ndash92 2018
[5] A Ghardallou A Kricha A Sakly and A Mtibaa ldquoAdaptiveblock sized reversible watermarking scheme based on integertransformrdquo in Proceedings of the 17th International Conferenceon Sciences and Techniques of Automatic Control and ComputerEngineering STA 2016 pp 347ndash351 Tunisia December 2016
[6] R G van Schyndel A Z Tirkel and C F Osborne ldquoA digitalwatermarkrdquo in Proceedings of the IEEE International ConferenceImage Processing (ICIP rsquo94) vol 2 pp 86ndash90 Austin Tex USANovember 1994
[7] W-H Lin S-J Horng T-W Kao P Fan C-L Lee and Y PanldquoAn efficient watermarking method based on significant differ-ence of wavelet coefficient quantizationrdquo IEEE Transactions onMultimedia vol 10 no 5 pp 746ndash757 2008
[8] S Bhattacharyya and G Sanyal ldquoA Robust Image Steganogra-phy using DWT Difference Modulation (DWTDM)rdquo Interna-tional Journal of Computer Network and Information Securityvol 4 no 7 pp 27ndash40 2012
[9] H S Malvar and D A Florencio ldquoImproved spread spectruma new modulation technique for robust watermarkingrdquo IEEETransactions on Signal Processing vol 51 no 4 pp 898ndash9052003
[10] B Chen and G W Wornell ldquoProvably robust digital water-markingrdquo in Proceedings of the 1999 Multimedia Systems andApplications II pp 43ndash54 September 1999
[11] Z Kricha A Kricha and A Sakly ldquoAccommodative extractorfor QIM-based watermarking schemesrdquo IET Image Processing2018
[12] I J Cox J Kilian T Leighton and T Shamoon ldquoSecure imper-ceptable yet perceptually salient spread spectrum watermark
formultimediardquo in Proceedings of the 1996 Southcon Conferencepp 192ndash197 June 1996
[13] X Li J Liu J Sun Z Sun H Hu and L Zhang ldquoImprovedspread transform dither modulation-based watermark algo-rithm based on step projectionrdquo in Proceedings of the 2010 IEEEInternational Symposium on Broadband Multimedia Systemsand Broadcasting (BMSB) pp 1ndash5 Shanghai China March2010
[14] H-T Hu and L-Y Hsu ldquoExploring DWT-SVD-DCT featureparameters for robust multiple watermarking against JPEG andJPEG2000 compressionrdquo Computers and Electrical Engineeringvol 41 pp 52ndash63 2015
[15] J Cannons and P Moulin ldquoDesign and statistical analysis of ahash-aided image watermarking systemrdquo IEEE Transactions onImage Processing vol 13 no 10 pp 1393ndash1408 2004
[16] A Valizadeh and Z J Wang ldquoAn improved multiplicativespread spectrum embedding scheme for data hidingrdquo IEEETransactions on Information Forensics and Security vol 7 no4 pp 1127ndash1143 2012
[17] M Kutter D Frederic and F B JordanDigital signature of colorimages using amplitude modulation 1997
[18] Q Su and B Chen ldquoRobust color image watermarking tech-nique in the spatial domainrdquo Soft Computing vol 22 no 1 pp91ndash106 2018
[19] Z Yanling Z Xiaoshi L Na et al ldquoA Digital ImageWatermarkAlgorithm Based onDCCoefficientsQuantizationrdquo in Proceed-ings of the 2006 6th World Congress on Intelligent Control andAutomation pp 9734ndash9738 Dalian China 2006
[20] RMZhaoH LianHWPang andBNHu ldquoAWatermarkingAlgorithm by Modifying AC Coefficies in DCT Domainrdquo inProceedings of the 2008 International Symposium on InformationScience and Engineering (ISISE) pp 159ndash162 Shanghai ChinaDecember 2008
[21] B L Gunjal and R RManthalkar ldquoDiscreteWavelet Transformbased strongly robust watermarking scheme for informationhiding in digital imagesrdquo in Proceedings of the 3rd InternationalConference on Emerging Trends in Engineering and TechnologyICETET 2010 pp 124ndash129 India November 2010
[22] H Shi and F Lv ldquoA Blind Digital Watermark Techniquefor Color Image Based on Integer Wavelet Transformrdquo inProceedings of the 2010 International Conference on BiomedicalEngineering and Computer Science (ICBECS) pp 1ndash4 WuhanChina April 2010
[23] S Ranjbar F Zargari and M Ghanbari ldquoA highly robust two-stage Contourlet-based digital image watermarking methodrdquoSignal Processing Image Communication vol 28 no 10 pp1526ndash1536 2013
[24] Q Su and B Chen ldquoA novel blind color image watermarkingusing upper Hessenbergmatrixrdquo AEU - International Journal ofElectronics and Communications vol 78 pp 64ndash71 2017
[25] F Ourique V Licks R Jordan and F Perez-Gonzalez ldquoAngleQIM a novel watermark embedding scheme robust againstamplitude scaling distortionsrdquo in Proceedings of the IEEE Inter-national Conference on Acoustics Speech and Signal Processing(ICASSP rsquo05) pp 797ndash800 March 2005
[26] H R Fazlali S Samavi N Karimi and S Shirani ldquoAdaptiveblind image watermarking using edge pixel concentrationrdquoMultimedia Tools and Applications vol 76 no 2 pp 3105ndash31202017
[27] X-Y Wang Y-N Liu H Xu A-L Wang and H-Y YangldquoBlind optimum detector for robust image watermarking in
16 Security and Communication Networks
nonsubsampled shearlet Domainrdquo Information Sciences vol372 pp 634ndash654 2016
[28] X Liu C Lin and S Yuan ldquoBlind Dual Watermarking forColor Imagesrsquo Authentication and Copyright Protectionrdquo IEEETransactions on Circuits and Systems for Video Technology vol28 no 5 pp 1047ndash1055 2018
[29] Poonam and S M Arora ldquoA DWT-SVD based Robust DigitalWatermarking for Digital Imagesrdquo Procedia Computer Sciencevol 132 pp 1441ndash1448 2018
[30] S Fazli and M Moeini ldquoA robust image watermarking methodbased on DWT DCT and SVD using a new technique forcorrection of main geometric attacksrdquo Optik - InternationalJournal for Light and ElectronOptics vol 127 no 2 pp 964ndash9722016
[31] K Kuppusamy and K Thamodaran ldquoOptimized Image Water-marking Scheme Based On PSOrdquo Procedia Engineering vol 38pp 493ndash503 2012
[32] V S Rao R S Shekhawat and V K Srivastava ldquoA dwt-dct-svd based digital image watermarking scheme using particleswarm optimizationrdquo in Proceedings of the 2012 IEEE StudentsrsquoConference on Electrical Electronics and Computer Science pp1ndash4 2012
[33] T T Takore P R Kumar and G L Devi ldquoA modified blindimage watermarking scheme based on DWT DCT and SVDdomain using GA to optimize robustnessrdquo in Proceedings ofthe 2016 International Conference on Electrical Electronics andOptimization Techniques ICEEOT 2016 pp 2725ndash2729 IndiaMarch 2016
[34] H Hu S Lin and L Hsu ldquoEffective blind speech watermarkingvia adaptive mean modulation and package synchronization inDWT domainrdquo EURASIP Journal on Audio Speech and MusicProcessing vol 2017 no 1 2017
[35] W-H Lin Y-R Wang S-J Horng T-W Kao and Y Pan ldquoAblind watermarking method using maximum wavelet coeffi-cient quantizationrdquo Expert Systems with Applications vol 36no 9 pp 11509ndash11516 2009
[36] X Li J Liu J Sun X Yang andW Liu ldquoStep-projection-basedspread transform dither modulationrdquo IET Information Securityvol 5 no 3 pp 170ndash180 2011
[37] M Hamghalam S Mirzakuchaki andM A Akhaee ldquoGeomet-ric modelling of the wavelet coefficients for image watermark-ing using optimum detectorrdquo IET Image Processing vol 8 no 3pp 162ndash172 2014
[38] V S Verma R K Jha and A Ojha ldquoDigital watermarkextraction using support vector machine with principal com-ponent analysis based feature reductionrdquo Journal of VisualCommunication and Image Representation vol 31 pp 75ndash852015
[39] C Li Z Zhang Y Wang B Ma and D Huang ldquoDithermodulation of significant amplitude difference for waveletbased robust watermarkingrdquo Neurocomputing vol 166 pp404ndash415 2015
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Navigation and Observation
International Journal of
Hindawi
wwwhindawicom Volume 2018
Advances in
Multimedia
Submit your manuscripts atwwwhindawicom
6 Security and Communication Networks
LL
HH
12
Chaotic sequence
4 2 7 1 6 5 3 8
12 09 -03 01 -24 -11 -07 -18
01 09 -07 12 -11 -24 -03 -18
022 025 028 067 074 081 090 0914 2 7 1 6 5 3 8
067 025 090 022 081 074 028 0911 2 3 4 5 6 7 8
Permutation sequence
Sorting09
-03 01
-24 -11
-07 -18C
C
Figure 5 Shuffling of DWT diagonal subbands using a chaotic sequence of eight real numbers
order (1199091199041 1199091199042 1199091199043 119909119904119899) and so the permutation sequence(1199041 1199042 1199043 119904119899) is obtained
The example shown in Figure 5 illustrates a concreteexample for the shuffling operation In this example 119899 is equalto eight the vector 119862 is the concatenation of the diagonalDWT coefficients and 119862 is the shuffled version of 11986233 Block Mean Modulation The first step in the blockmodulation phase is the partitioning of the confused hostcoefficients 119888119894 into independent blocks This operation isrealized by gathering each 119905 adjacent coefficients into a singleblock 119861119871 119894 as given by (2) where 119894 is the index of the blockand is the concatenation symbol The total number ofgenerated blocks 119861119871 should be larger or equal than the binarywatermark message length In other words lfloor119899119905rfloor ge 119898 suchas 119905 isin Nlowast
To embed a robust watermark we consider the quantizationtechnique which modulates the carrier hosts to 119875 nonover-lapping sets where 119875 represents the possible embeddedwords In this work the watermark is a binary messagehence the carrier will be modulated to two different sets(119901 = 2) each associated to a watermark bit The robustnessof the quantization-based insertion method depends on thedistance between the two sets On the other hand wideningthe gap will lead tomediocre results in terms of visual qualityTo work around this issue Chen et al [10] proposed toincrease the element of each set and it is up to the embeddingalgorithm to choose the appropriate quantizer among theavailable ones However as illustrated in Figure 3(a) mostof the BMV are concentrated around the zero where in 80of the BMVs are in the range [minus20 20] Thus we propose tomodulate the blocksrsquo mean values by a positive or negativenumber according to the embedded bit Let 119861119871 119894 be the meanvalue of the block 119894 defined by
119861119871 119894 = 1119905119894times119905sum
119895=119905(119894minus1)+1
119888119895 (3)
The modulation of 119861119871 119894 goes by firstly nullifying the meanvalue by spreading its value equally to all blockrsquos coefficientsthen adding or subtracting a threshold 119879 according to theembedded bit 119887 The equation of modulation is given by (4)where 119888119896 is a coefficient belonging to the block 119894 (119861119871 119894)
1198881015840119896 = 119888119896 minus 119861119871 119894119905 minus 119879 if 119887 = 0119888119896 minus 119861119871 119894119905 + 119879 otherwise (4)
Once all the watermark bits are inserted the watermarkedcoefficients are rearranged into their initial positions usingthe same key SK Finally the inverse DWT is performed toobtain the watermarked image
34 Watermark Extraction In the watermark extractionphase the watermarked image is firstly transformed in thefrequency domain as in the embedding and the same key SKis used to constitute different blocks The decision upon theembedded bit is judged according to the blockrsquos mean valueof the attacked watermarked block 119861119871119894
119887 = 1 if 119861119871119894 gt 00 otherwise (5)
4 Experimental Results
This section is dedicated to evaluate the performance ofthe suggested watermarking framework A variety of 8-bitdepth (119889 = 8) grayscale standard benchmark images witha resolution of 512 times 512 pixels (119908 = ℎ = 256) have beensubjects of assessment The used image database includesLena Baboon Peppers and Tank These images present avariety of image properties (eg mixture of dark and lightarea alongside with highly textures regions (hair) in the Lenaimage) which refutes the idea of tying the performance of thescheme to some image properties
The performance of the suggested method is experimen-tally examined from different aspects Firstly the relation
Security and Communication Networks 7
38 41 45PSNR (dB)
2
4
6
8
10
12
14
16
18
Thre
shol
d va
lue (
T)
123
DWT level
Figure 6 Relation of threshold value (T) and the PSNR at different DWT levels
between the capacity and the imperceptibility is evaluatedusing different insertion parameters Next the evolution ofcomplexity according to the host image size is monitoredAfterwards the robustness is validated in two ways initiallythe robustness against different levels of attacks is examinedand then a comparative study with similar works is presentedFinally the security of the scheme is verified
41 Imperceptibility The imperceptibility feature reflects thefidelity of the watermarked image toward the original oneSeveral metrics can be used to quantify the noticeable differ-ence between two images involving subjective and objectivemetrics In this work we employed the Peek Signal to NoiseRatio (PSNR) and the Structural Similarity Index Measure(SSIM) given by (6) and (7) respectively
PSNR (I Ilowast) = 10 log10( (2119889 minus 1)2 times 119908 times ℎsum119908119894=1sumℎ119894=1 (I119894119895 minus Ilowast119894119895)2) (6)
where 120579I and 120583I are respectively the mean intensity and thestandard deviation of the original image I and 120590119868119868lowast is thesample cross-correlation between I and Ilowast The constants 1198781and 1198782 are defined as 1198781 = (1198961(2119889minus1))2 and 1198782 = (1198962(2119889minus1))2where 1198961 1198962 ≪ 142 Capacity The maximum capacity of the proposedscheme depends on three variables the image dimension (119908
and ℎ) the wavelet level (119897) and the block length (119905) Hencewe can express the maximum capacity using
Max capacity = lfloor 119908 times ℎ119905 times 22119897minus3rfloor (8)
For a better understanding of the relation between the capac-ity and imperceptibility we embedded a variable message size(119898 isin 256 512 1024) at different levels of decomposition(119897 isin 1 2 3) into three different images namely Lena Coupleand Baboon For each watermarking operation the blocksrsquosize 119905 is maximized so that all coefficients are involved inthe watermarking process (eg 119897 = 2 and 119905 = 32 for119898 = 1024) The threshold 119879 is manually chosen to reach aPSNR value of 38 41 and 45 dBThe parameters of insertion(119879 119905 119897 119898) the objective measurements (PSNR and SSIM)and the watermarked images are all listed in Table 1
The first observation that we draw from Table 1 is thatfor a given PSNR the threshold 119879 is converging towardthe same value regardless the host image or the watermarksize This amounts to saying that we can anticipate the119879 value according to the level of transformation and thetargeted PSNR Figure 6 illustrates the relation between theseparameters where119879 is obtained by averaging all the values fora specific level and PSNR Secondly it is shown from Table 1that at 45 dB the distortion due to the watermark insertionis invisible whatever the size of the message and the levelof decomposition However for watermarked images with aPSNR of 38 and 41 dB the imperceptibility is conditionedupon the level 119897 the higher 119897 is the lower the quality is goingto be
43 Complexity The computational cost of the watermarkembedding and extraction is an important aspect that may
8 Security and Communication Networks
Table 1 Quality of watermarked images using different insertion parameters (119897 119879119898)Message size Transformation
level Watermarked Lena images Watermarked Coupleimages
Figure 7 Computational time of (a) overall insertion and detection and (b) details of the embedding stage
restrict or expand the use of a given watermarking schemeIn the literature most of schemes are evaluated in a standardimage with a size of 512 times 512 pixels and there is noperspective of the computational cost trend when the imagegoes bigger which is more realistic Hence we propose tovary the size of the image beginning with 64 times 64 pixelsand ending with 4096 times 4096 pixels by doubling the heightand width after each insertion For the sake of simplicity thewatermark was embedded after one level of DWT (119897 = 1)with a fixed block size (119905 = 4) and a maximum watermarksize as defined by (8)
Figure 7(a) depicts the execution time evolution ofthe overall embedding and detection alongside with thevariation of the watermark size (red) as a function of theimage size The details of the embedding phase are shownin Figure 7(b) The most important remark to be made fromFigure 7 is that the computational time evolves broadly inline with the watermark size (red curve) to find out thecomplexity of our scheme in comparison with other state-ofthe-art methods
44 Robustness In the following we investigate the robust-ness of the suggested watermarking approach under severaltypes of attacks The robustness is quantified using the BitError Rate (BER) and the Normalized Correlation (NC)metrics
441 Evaluated Attacks Since the proposed scheme isdesigned to be robust the embedded message should lastor at list a significant part of it after image processingoperations Fromwatermark perspective all image processingoperations whatever performed on goodwill or not is consid-ered as attacks The involved attacks in the evaluation tests areclassified into three groups
(i) Compression attacks usually images are not storedin a raw format but after a compression operationthat aims to reduce the required storage space JPEGalgorithm is one of the most used compression
algorithms However JPEG is a lossy algorithm thatcauses some quality degradation inversely commen-surate with the compression quality factor Hence wetested the robustness of our algorithm under differentJPEG quality factors ranging from 10 to 90
(ii) Image processing attacks used to imitate a noisecontaminating the image during transmission such assalt and peppers or to improve the imagersquos qualityby eliminating a noise or enhancing some propertiesThe used attacks are listed as follows scaling medianfiltering sharpening Gaussian filtering average fil-tering Gaussian noise salt and pepper noise contam-ination histogram equalization and additive whiteGaussian noise (AWGN)
(iii) Geometric attacks used to correct geometric distor-tions such camera orientation (rotation) or imagesrsquoalignment (cropping) The used attacks are listed asfollows image cropping and resizing and rotation
Since the DWT coefficients are not geometrically invari-ant a manual synchronization of the imagersquos dimension andorientation was realized prior to the extraction
442 Performance Evaluation Here we test the robustnessof our scheme on three different images Boat Peppersand Barbara Each time we varied the embedded messagebetween 128 bits and 2048 bits The embedding threshold 119879is manually tuned to provide a watermarked image with aPSNR value of 40 dB and 45 dB The robustness is expressedin terms of BER () after mean filter (Figure 8) JPEGcompression (Figure 9) median filter (Figure 10) and gammacorrection Poisson noise histogram equalization and imageunsharpening (Figure 11) Figures 8ndash11 show that the robust-ness performance slightly differs between images For a smallmessage length (between 128 and 256 bits) and a PSNR valueof 40 dB the BER does not exceed 20 for most of attacksHowever the larger the message size is or the stronger theattack is the higher the BER will be
10 Security and Communication Networks
128 256 512 1024 2048Watermark size
0
10
20
30
40
50BE
R (
)
Mean filter (3x3)Mean filter (5x5)Mean filter (7x7)Mean filter (9x9)BarbaraPeppersBoat
(a)
Mean filter (3x3)Mean filter (5x5)Mean filter (7x7)Mean filter (9x9)BarbaraPeppersBoat
128 256 512 1024 2048Watermark size
0
10
20
30
40
50
BER
()
(b)
Figure 8 BER() after mean filter attacks using different message lengths at different PSNR values (a) 40 dB and (b) 45 dB
Figure 9 BER() after JPEG compression using different message lengths at different PSNR values (a) 40 dB and (b) 45 dB
443 Comparative Study To prove the superiority and therobustness advantages of our scheme a comparative studywith state of the art works is indispensable In fact wecompared the robustness of our scheme with six robustwatermarking schemes [26 27 35 37ndash39] In order to carryout a fair comparison the same insertion parameters (PSNRwatermark size) and attacks are used in this comparisonThe PSNRs value of different images and their correspondingvalues in the comparative schemes are illustrated in Figure 12It is worth mentioning that the insertion parameters (119897 119905 119879)were empirically chosen to produce the same PSNR alongsidewith an acceptable visual quality
The robustness results are presented in Tables 2ndash5 interms of either BER or NC depending on how they appearin the reference papers and the superior results are markedin bold number
Based on the simulation results given in Tables 2ndash5 it canbe found that our approach outperforms similar schemes on7194 of the time Equal results are obtained on 2343 ofthe time and our scheme underperforms on 462 of thecases
45 Security The robustness of a watermarking scheme isnot sufficient to judge a scheme as reliable As discussedin the first section robust watermarking is used to secureimages Hence the security is necessary First the watermarkshould be immune against illegal extraction even when theembedding procedure is exposed Second the watermarkdetector should not report the existence of the watermarkwhen none is embedded To evaluate these two risks weextract the watermark from a watermarked image using
Security and Communication Networks 11
Table 2 BER comparison with [27] and [37] using a message length of 128 bits
different keys and extract a watermark from a nonwater-marked image respectively In the two cases the maximumcorrelation obtained has not exceeded the value 059 whichpoints out a negative presence
5 Conclusion
The mean value of DWTrsquos vertical and horizontal subbandscoefficients is highly robust and imperceptible In fact these
two advantages were exploited in the design of a robustand blind watermarking approach The embedding methodconsisted in modulating the BMV in accordance with thequantization technique The use of a single transformationdomain alongside with a simple embedding technique makesfrom the proposed algorithmcomputationally undemandingAt the extraction phase the watermark bits were identifiedby the BMV position with respect to the zero value Unlikesome existing schemes the extraction rules did not require to
Security and Communication Networks 15
be adjusted to the applied attacks to fulfill a high robustnessIn fact the inserted watermark was capable to withstand withno error after JPEG compression up to a quality factor of 30Gaussian filter and other attacks
Data Availability
The data used to support the findings of this study areincluded within the article
Conflicts of Interest
The authors declare that they have no conflicts of interest
References
[1] S Liansheng Z Bei W Zhanmin and T Ailing ldquoAn opticalcolor image watermarking scheme by using compressive sens-ingwith human visual characteristics in gyrator domainrdquoOpticsand Lasers in Engineering vol 92 pp 85ndash93 2017
[2] D Shukla and M Sharma ldquoA Novel Scene-Based Video Water-marking Scheme for Copyright Protectionrdquo Journal of IntelligentSystems vol 27 no 1 pp 47ndash66 2018
[3] A Tiwari and M Sharma ldquoAn efficient vector quantizationbased watermarking method for image integrity authentica-tionrdquo Advances in Intelligent Systems and Computing vol 519pp 215ndash225 2018
[4] S Bravo-Solorio F Calderon C-T Li and A K Nandi ldquoFastfragile watermark embedding and iterative mechanism withhigh self-restoration performancerdquo Digital Signal Processingvol 73 pp 83ndash92 2018
[5] A Ghardallou A Kricha A Sakly and A Mtibaa ldquoAdaptiveblock sized reversible watermarking scheme based on integertransformrdquo in Proceedings of the 17th International Conferenceon Sciences and Techniques of Automatic Control and ComputerEngineering STA 2016 pp 347ndash351 Tunisia December 2016
[6] R G van Schyndel A Z Tirkel and C F Osborne ldquoA digitalwatermarkrdquo in Proceedings of the IEEE International ConferenceImage Processing (ICIP rsquo94) vol 2 pp 86ndash90 Austin Tex USANovember 1994
[7] W-H Lin S-J Horng T-W Kao P Fan C-L Lee and Y PanldquoAn efficient watermarking method based on significant differ-ence of wavelet coefficient quantizationrdquo IEEE Transactions onMultimedia vol 10 no 5 pp 746ndash757 2008
[8] S Bhattacharyya and G Sanyal ldquoA Robust Image Steganogra-phy using DWT Difference Modulation (DWTDM)rdquo Interna-tional Journal of Computer Network and Information Securityvol 4 no 7 pp 27ndash40 2012
[9] H S Malvar and D A Florencio ldquoImproved spread spectruma new modulation technique for robust watermarkingrdquo IEEETransactions on Signal Processing vol 51 no 4 pp 898ndash9052003
[10] B Chen and G W Wornell ldquoProvably robust digital water-markingrdquo in Proceedings of the 1999 Multimedia Systems andApplications II pp 43ndash54 September 1999
[11] Z Kricha A Kricha and A Sakly ldquoAccommodative extractorfor QIM-based watermarking schemesrdquo IET Image Processing2018
[12] I J Cox J Kilian T Leighton and T Shamoon ldquoSecure imper-ceptable yet perceptually salient spread spectrum watermark
formultimediardquo in Proceedings of the 1996 Southcon Conferencepp 192ndash197 June 1996
[13] X Li J Liu J Sun Z Sun H Hu and L Zhang ldquoImprovedspread transform dither modulation-based watermark algo-rithm based on step projectionrdquo in Proceedings of the 2010 IEEEInternational Symposium on Broadband Multimedia Systemsand Broadcasting (BMSB) pp 1ndash5 Shanghai China March2010
[14] H-T Hu and L-Y Hsu ldquoExploring DWT-SVD-DCT featureparameters for robust multiple watermarking against JPEG andJPEG2000 compressionrdquo Computers and Electrical Engineeringvol 41 pp 52ndash63 2015
[15] J Cannons and P Moulin ldquoDesign and statistical analysis of ahash-aided image watermarking systemrdquo IEEE Transactions onImage Processing vol 13 no 10 pp 1393ndash1408 2004
[16] A Valizadeh and Z J Wang ldquoAn improved multiplicativespread spectrum embedding scheme for data hidingrdquo IEEETransactions on Information Forensics and Security vol 7 no4 pp 1127ndash1143 2012
[17] M Kutter D Frederic and F B JordanDigital signature of colorimages using amplitude modulation 1997
[18] Q Su and B Chen ldquoRobust color image watermarking tech-nique in the spatial domainrdquo Soft Computing vol 22 no 1 pp91ndash106 2018
[19] Z Yanling Z Xiaoshi L Na et al ldquoA Digital ImageWatermarkAlgorithm Based onDCCoefficientsQuantizationrdquo in Proceed-ings of the 2006 6th World Congress on Intelligent Control andAutomation pp 9734ndash9738 Dalian China 2006
[20] RMZhaoH LianHWPang andBNHu ldquoAWatermarkingAlgorithm by Modifying AC Coefficies in DCT Domainrdquo inProceedings of the 2008 International Symposium on InformationScience and Engineering (ISISE) pp 159ndash162 Shanghai ChinaDecember 2008
[21] B L Gunjal and R RManthalkar ldquoDiscreteWavelet Transformbased strongly robust watermarking scheme for informationhiding in digital imagesrdquo in Proceedings of the 3rd InternationalConference on Emerging Trends in Engineering and TechnologyICETET 2010 pp 124ndash129 India November 2010
[22] H Shi and F Lv ldquoA Blind Digital Watermark Techniquefor Color Image Based on Integer Wavelet Transformrdquo inProceedings of the 2010 International Conference on BiomedicalEngineering and Computer Science (ICBECS) pp 1ndash4 WuhanChina April 2010
[23] S Ranjbar F Zargari and M Ghanbari ldquoA highly robust two-stage Contourlet-based digital image watermarking methodrdquoSignal Processing Image Communication vol 28 no 10 pp1526ndash1536 2013
[24] Q Su and B Chen ldquoA novel blind color image watermarkingusing upper Hessenbergmatrixrdquo AEU - International Journal ofElectronics and Communications vol 78 pp 64ndash71 2017
[25] F Ourique V Licks R Jordan and F Perez-Gonzalez ldquoAngleQIM a novel watermark embedding scheme robust againstamplitude scaling distortionsrdquo in Proceedings of the IEEE Inter-national Conference on Acoustics Speech and Signal Processing(ICASSP rsquo05) pp 797ndash800 March 2005
[26] H R Fazlali S Samavi N Karimi and S Shirani ldquoAdaptiveblind image watermarking using edge pixel concentrationrdquoMultimedia Tools and Applications vol 76 no 2 pp 3105ndash31202017
[27] X-Y Wang Y-N Liu H Xu A-L Wang and H-Y YangldquoBlind optimum detector for robust image watermarking in
16 Security and Communication Networks
nonsubsampled shearlet Domainrdquo Information Sciences vol372 pp 634ndash654 2016
[28] X Liu C Lin and S Yuan ldquoBlind Dual Watermarking forColor Imagesrsquo Authentication and Copyright Protectionrdquo IEEETransactions on Circuits and Systems for Video Technology vol28 no 5 pp 1047ndash1055 2018
[29] Poonam and S M Arora ldquoA DWT-SVD based Robust DigitalWatermarking for Digital Imagesrdquo Procedia Computer Sciencevol 132 pp 1441ndash1448 2018
[30] S Fazli and M Moeini ldquoA robust image watermarking methodbased on DWT DCT and SVD using a new technique forcorrection of main geometric attacksrdquo Optik - InternationalJournal for Light and ElectronOptics vol 127 no 2 pp 964ndash9722016
[31] K Kuppusamy and K Thamodaran ldquoOptimized Image Water-marking Scheme Based On PSOrdquo Procedia Engineering vol 38pp 493ndash503 2012
[32] V S Rao R S Shekhawat and V K Srivastava ldquoA dwt-dct-svd based digital image watermarking scheme using particleswarm optimizationrdquo in Proceedings of the 2012 IEEE StudentsrsquoConference on Electrical Electronics and Computer Science pp1ndash4 2012
[33] T T Takore P R Kumar and G L Devi ldquoA modified blindimage watermarking scheme based on DWT DCT and SVDdomain using GA to optimize robustnessrdquo in Proceedings ofthe 2016 International Conference on Electrical Electronics andOptimization Techniques ICEEOT 2016 pp 2725ndash2729 IndiaMarch 2016
[34] H Hu S Lin and L Hsu ldquoEffective blind speech watermarkingvia adaptive mean modulation and package synchronization inDWT domainrdquo EURASIP Journal on Audio Speech and MusicProcessing vol 2017 no 1 2017
[35] W-H Lin Y-R Wang S-J Horng T-W Kao and Y Pan ldquoAblind watermarking method using maximum wavelet coeffi-cient quantizationrdquo Expert Systems with Applications vol 36no 9 pp 11509ndash11516 2009
[36] X Li J Liu J Sun X Yang andW Liu ldquoStep-projection-basedspread transform dither modulationrdquo IET Information Securityvol 5 no 3 pp 170ndash180 2011
[37] M Hamghalam S Mirzakuchaki andM A Akhaee ldquoGeomet-ric modelling of the wavelet coefficients for image watermark-ing using optimum detectorrdquo IET Image Processing vol 8 no 3pp 162ndash172 2014
[38] V S Verma R K Jha and A Ojha ldquoDigital watermarkextraction using support vector machine with principal com-ponent analysis based feature reductionrdquo Journal of VisualCommunication and Image Representation vol 31 pp 75ndash852015
[39] C Li Z Zhang Y Wang B Ma and D Huang ldquoDithermodulation of significant amplitude difference for waveletbased robust watermarkingrdquo Neurocomputing vol 166 pp404ndash415 2015
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Navigation and Observation
International Journal of
Hindawi
wwwhindawicom Volume 2018
Advances in
Multimedia
Submit your manuscripts atwwwhindawicom
Security and Communication Networks 7
38 41 45PSNR (dB)
2
4
6
8
10
12
14
16
18
Thre
shol
d va
lue (
T)
123
DWT level
Figure 6 Relation of threshold value (T) and the PSNR at different DWT levels
between the capacity and the imperceptibility is evaluatedusing different insertion parameters Next the evolution ofcomplexity according to the host image size is monitoredAfterwards the robustness is validated in two ways initiallythe robustness against different levels of attacks is examinedand then a comparative study with similar works is presentedFinally the security of the scheme is verified
41 Imperceptibility The imperceptibility feature reflects thefidelity of the watermarked image toward the original oneSeveral metrics can be used to quantify the noticeable differ-ence between two images involving subjective and objectivemetrics In this work we employed the Peek Signal to NoiseRatio (PSNR) and the Structural Similarity Index Measure(SSIM) given by (6) and (7) respectively
PSNR (I Ilowast) = 10 log10( (2119889 minus 1)2 times 119908 times ℎsum119908119894=1sumℎ119894=1 (I119894119895 minus Ilowast119894119895)2) (6)
where 120579I and 120583I are respectively the mean intensity and thestandard deviation of the original image I and 120590119868119868lowast is thesample cross-correlation between I and Ilowast The constants 1198781and 1198782 are defined as 1198781 = (1198961(2119889minus1))2 and 1198782 = (1198962(2119889minus1))2where 1198961 1198962 ≪ 142 Capacity The maximum capacity of the proposedscheme depends on three variables the image dimension (119908
and ℎ) the wavelet level (119897) and the block length (119905) Hencewe can express the maximum capacity using
Max capacity = lfloor 119908 times ℎ119905 times 22119897minus3rfloor (8)
For a better understanding of the relation between the capac-ity and imperceptibility we embedded a variable message size(119898 isin 256 512 1024) at different levels of decomposition(119897 isin 1 2 3) into three different images namely Lena Coupleand Baboon For each watermarking operation the blocksrsquosize 119905 is maximized so that all coefficients are involved inthe watermarking process (eg 119897 = 2 and 119905 = 32 for119898 = 1024) The threshold 119879 is manually chosen to reach aPSNR value of 38 41 and 45 dBThe parameters of insertion(119879 119905 119897 119898) the objective measurements (PSNR and SSIM)and the watermarked images are all listed in Table 1
The first observation that we draw from Table 1 is thatfor a given PSNR the threshold 119879 is converging towardthe same value regardless the host image or the watermarksize This amounts to saying that we can anticipate the119879 value according to the level of transformation and thetargeted PSNR Figure 6 illustrates the relation between theseparameters where119879 is obtained by averaging all the values fora specific level and PSNR Secondly it is shown from Table 1that at 45 dB the distortion due to the watermark insertionis invisible whatever the size of the message and the levelof decomposition However for watermarked images with aPSNR of 38 and 41 dB the imperceptibility is conditionedupon the level 119897 the higher 119897 is the lower the quality is goingto be
43 Complexity The computational cost of the watermarkembedding and extraction is an important aspect that may
8 Security and Communication Networks
Table 1 Quality of watermarked images using different insertion parameters (119897 119879119898)Message size Transformation
level Watermarked Lena images Watermarked Coupleimages
Figure 7 Computational time of (a) overall insertion and detection and (b) details of the embedding stage
restrict or expand the use of a given watermarking schemeIn the literature most of schemes are evaluated in a standardimage with a size of 512 times 512 pixels and there is noperspective of the computational cost trend when the imagegoes bigger which is more realistic Hence we propose tovary the size of the image beginning with 64 times 64 pixelsand ending with 4096 times 4096 pixels by doubling the heightand width after each insertion For the sake of simplicity thewatermark was embedded after one level of DWT (119897 = 1)with a fixed block size (119905 = 4) and a maximum watermarksize as defined by (8)
Figure 7(a) depicts the execution time evolution ofthe overall embedding and detection alongside with thevariation of the watermark size (red) as a function of theimage size The details of the embedding phase are shownin Figure 7(b) The most important remark to be made fromFigure 7 is that the computational time evolves broadly inline with the watermark size (red curve) to find out thecomplexity of our scheme in comparison with other state-ofthe-art methods
44 Robustness In the following we investigate the robust-ness of the suggested watermarking approach under severaltypes of attacks The robustness is quantified using the BitError Rate (BER) and the Normalized Correlation (NC)metrics
441 Evaluated Attacks Since the proposed scheme isdesigned to be robust the embedded message should lastor at list a significant part of it after image processingoperations Fromwatermark perspective all image processingoperations whatever performed on goodwill or not is consid-ered as attacks The involved attacks in the evaluation tests areclassified into three groups
(i) Compression attacks usually images are not storedin a raw format but after a compression operationthat aims to reduce the required storage space JPEGalgorithm is one of the most used compression
algorithms However JPEG is a lossy algorithm thatcauses some quality degradation inversely commen-surate with the compression quality factor Hence wetested the robustness of our algorithm under differentJPEG quality factors ranging from 10 to 90
(ii) Image processing attacks used to imitate a noisecontaminating the image during transmission such assalt and peppers or to improve the imagersquos qualityby eliminating a noise or enhancing some propertiesThe used attacks are listed as follows scaling medianfiltering sharpening Gaussian filtering average fil-tering Gaussian noise salt and pepper noise contam-ination histogram equalization and additive whiteGaussian noise (AWGN)
(iii) Geometric attacks used to correct geometric distor-tions such camera orientation (rotation) or imagesrsquoalignment (cropping) The used attacks are listed asfollows image cropping and resizing and rotation
Since the DWT coefficients are not geometrically invari-ant a manual synchronization of the imagersquos dimension andorientation was realized prior to the extraction
442 Performance Evaluation Here we test the robustnessof our scheme on three different images Boat Peppersand Barbara Each time we varied the embedded messagebetween 128 bits and 2048 bits The embedding threshold 119879is manually tuned to provide a watermarked image with aPSNR value of 40 dB and 45 dB The robustness is expressedin terms of BER () after mean filter (Figure 8) JPEGcompression (Figure 9) median filter (Figure 10) and gammacorrection Poisson noise histogram equalization and imageunsharpening (Figure 11) Figures 8ndash11 show that the robust-ness performance slightly differs between images For a smallmessage length (between 128 and 256 bits) and a PSNR valueof 40 dB the BER does not exceed 20 for most of attacksHowever the larger the message size is or the stronger theattack is the higher the BER will be
10 Security and Communication Networks
128 256 512 1024 2048Watermark size
0
10
20
30
40
50BE
R (
)
Mean filter (3x3)Mean filter (5x5)Mean filter (7x7)Mean filter (9x9)BarbaraPeppersBoat
(a)
Mean filter (3x3)Mean filter (5x5)Mean filter (7x7)Mean filter (9x9)BarbaraPeppersBoat
128 256 512 1024 2048Watermark size
0
10
20
30
40
50
BER
()
(b)
Figure 8 BER() after mean filter attacks using different message lengths at different PSNR values (a) 40 dB and (b) 45 dB
Figure 9 BER() after JPEG compression using different message lengths at different PSNR values (a) 40 dB and (b) 45 dB
443 Comparative Study To prove the superiority and therobustness advantages of our scheme a comparative studywith state of the art works is indispensable In fact wecompared the robustness of our scheme with six robustwatermarking schemes [26 27 35 37ndash39] In order to carryout a fair comparison the same insertion parameters (PSNRwatermark size) and attacks are used in this comparisonThe PSNRs value of different images and their correspondingvalues in the comparative schemes are illustrated in Figure 12It is worth mentioning that the insertion parameters (119897 119905 119879)were empirically chosen to produce the same PSNR alongsidewith an acceptable visual quality
The robustness results are presented in Tables 2ndash5 interms of either BER or NC depending on how they appearin the reference papers and the superior results are markedin bold number
Based on the simulation results given in Tables 2ndash5 it canbe found that our approach outperforms similar schemes on7194 of the time Equal results are obtained on 2343 ofthe time and our scheme underperforms on 462 of thecases
45 Security The robustness of a watermarking scheme isnot sufficient to judge a scheme as reliable As discussedin the first section robust watermarking is used to secureimages Hence the security is necessary First the watermarkshould be immune against illegal extraction even when theembedding procedure is exposed Second the watermarkdetector should not report the existence of the watermarkwhen none is embedded To evaluate these two risks weextract the watermark from a watermarked image using
Security and Communication Networks 11
Table 2 BER comparison with [27] and [37] using a message length of 128 bits
different keys and extract a watermark from a nonwater-marked image respectively In the two cases the maximumcorrelation obtained has not exceeded the value 059 whichpoints out a negative presence
5 Conclusion
The mean value of DWTrsquos vertical and horizontal subbandscoefficients is highly robust and imperceptible In fact these
two advantages were exploited in the design of a robustand blind watermarking approach The embedding methodconsisted in modulating the BMV in accordance with thequantization technique The use of a single transformationdomain alongside with a simple embedding technique makesfrom the proposed algorithmcomputationally undemandingAt the extraction phase the watermark bits were identifiedby the BMV position with respect to the zero value Unlikesome existing schemes the extraction rules did not require to
Security and Communication Networks 15
be adjusted to the applied attacks to fulfill a high robustnessIn fact the inserted watermark was capable to withstand withno error after JPEG compression up to a quality factor of 30Gaussian filter and other attacks
Data Availability
The data used to support the findings of this study areincluded within the article
Conflicts of Interest
The authors declare that they have no conflicts of interest
References
[1] S Liansheng Z Bei W Zhanmin and T Ailing ldquoAn opticalcolor image watermarking scheme by using compressive sens-ingwith human visual characteristics in gyrator domainrdquoOpticsand Lasers in Engineering vol 92 pp 85ndash93 2017
[2] D Shukla and M Sharma ldquoA Novel Scene-Based Video Water-marking Scheme for Copyright Protectionrdquo Journal of IntelligentSystems vol 27 no 1 pp 47ndash66 2018
[3] A Tiwari and M Sharma ldquoAn efficient vector quantizationbased watermarking method for image integrity authentica-tionrdquo Advances in Intelligent Systems and Computing vol 519pp 215ndash225 2018
[4] S Bravo-Solorio F Calderon C-T Li and A K Nandi ldquoFastfragile watermark embedding and iterative mechanism withhigh self-restoration performancerdquo Digital Signal Processingvol 73 pp 83ndash92 2018
[5] A Ghardallou A Kricha A Sakly and A Mtibaa ldquoAdaptiveblock sized reversible watermarking scheme based on integertransformrdquo in Proceedings of the 17th International Conferenceon Sciences and Techniques of Automatic Control and ComputerEngineering STA 2016 pp 347ndash351 Tunisia December 2016
[6] R G van Schyndel A Z Tirkel and C F Osborne ldquoA digitalwatermarkrdquo in Proceedings of the IEEE International ConferenceImage Processing (ICIP rsquo94) vol 2 pp 86ndash90 Austin Tex USANovember 1994
[7] W-H Lin S-J Horng T-W Kao P Fan C-L Lee and Y PanldquoAn efficient watermarking method based on significant differ-ence of wavelet coefficient quantizationrdquo IEEE Transactions onMultimedia vol 10 no 5 pp 746ndash757 2008
[8] S Bhattacharyya and G Sanyal ldquoA Robust Image Steganogra-phy using DWT Difference Modulation (DWTDM)rdquo Interna-tional Journal of Computer Network and Information Securityvol 4 no 7 pp 27ndash40 2012
[9] H S Malvar and D A Florencio ldquoImproved spread spectruma new modulation technique for robust watermarkingrdquo IEEETransactions on Signal Processing vol 51 no 4 pp 898ndash9052003
[10] B Chen and G W Wornell ldquoProvably robust digital water-markingrdquo in Proceedings of the 1999 Multimedia Systems andApplications II pp 43ndash54 September 1999
[11] Z Kricha A Kricha and A Sakly ldquoAccommodative extractorfor QIM-based watermarking schemesrdquo IET Image Processing2018
[12] I J Cox J Kilian T Leighton and T Shamoon ldquoSecure imper-ceptable yet perceptually salient spread spectrum watermark
formultimediardquo in Proceedings of the 1996 Southcon Conferencepp 192ndash197 June 1996
[13] X Li J Liu J Sun Z Sun H Hu and L Zhang ldquoImprovedspread transform dither modulation-based watermark algo-rithm based on step projectionrdquo in Proceedings of the 2010 IEEEInternational Symposium on Broadband Multimedia Systemsand Broadcasting (BMSB) pp 1ndash5 Shanghai China March2010
[14] H-T Hu and L-Y Hsu ldquoExploring DWT-SVD-DCT featureparameters for robust multiple watermarking against JPEG andJPEG2000 compressionrdquo Computers and Electrical Engineeringvol 41 pp 52ndash63 2015
[15] J Cannons and P Moulin ldquoDesign and statistical analysis of ahash-aided image watermarking systemrdquo IEEE Transactions onImage Processing vol 13 no 10 pp 1393ndash1408 2004
[16] A Valizadeh and Z J Wang ldquoAn improved multiplicativespread spectrum embedding scheme for data hidingrdquo IEEETransactions on Information Forensics and Security vol 7 no4 pp 1127ndash1143 2012
[17] M Kutter D Frederic and F B JordanDigital signature of colorimages using amplitude modulation 1997
[18] Q Su and B Chen ldquoRobust color image watermarking tech-nique in the spatial domainrdquo Soft Computing vol 22 no 1 pp91ndash106 2018
[19] Z Yanling Z Xiaoshi L Na et al ldquoA Digital ImageWatermarkAlgorithm Based onDCCoefficientsQuantizationrdquo in Proceed-ings of the 2006 6th World Congress on Intelligent Control andAutomation pp 9734ndash9738 Dalian China 2006
[20] RMZhaoH LianHWPang andBNHu ldquoAWatermarkingAlgorithm by Modifying AC Coefficies in DCT Domainrdquo inProceedings of the 2008 International Symposium on InformationScience and Engineering (ISISE) pp 159ndash162 Shanghai ChinaDecember 2008
[21] B L Gunjal and R RManthalkar ldquoDiscreteWavelet Transformbased strongly robust watermarking scheme for informationhiding in digital imagesrdquo in Proceedings of the 3rd InternationalConference on Emerging Trends in Engineering and TechnologyICETET 2010 pp 124ndash129 India November 2010
[22] H Shi and F Lv ldquoA Blind Digital Watermark Techniquefor Color Image Based on Integer Wavelet Transformrdquo inProceedings of the 2010 International Conference on BiomedicalEngineering and Computer Science (ICBECS) pp 1ndash4 WuhanChina April 2010
[23] S Ranjbar F Zargari and M Ghanbari ldquoA highly robust two-stage Contourlet-based digital image watermarking methodrdquoSignal Processing Image Communication vol 28 no 10 pp1526ndash1536 2013
[24] Q Su and B Chen ldquoA novel blind color image watermarkingusing upper Hessenbergmatrixrdquo AEU - International Journal ofElectronics and Communications vol 78 pp 64ndash71 2017
[25] F Ourique V Licks R Jordan and F Perez-Gonzalez ldquoAngleQIM a novel watermark embedding scheme robust againstamplitude scaling distortionsrdquo in Proceedings of the IEEE Inter-national Conference on Acoustics Speech and Signal Processing(ICASSP rsquo05) pp 797ndash800 March 2005
[26] H R Fazlali S Samavi N Karimi and S Shirani ldquoAdaptiveblind image watermarking using edge pixel concentrationrdquoMultimedia Tools and Applications vol 76 no 2 pp 3105ndash31202017
[27] X-Y Wang Y-N Liu H Xu A-L Wang and H-Y YangldquoBlind optimum detector for robust image watermarking in
16 Security and Communication Networks
nonsubsampled shearlet Domainrdquo Information Sciences vol372 pp 634ndash654 2016
[28] X Liu C Lin and S Yuan ldquoBlind Dual Watermarking forColor Imagesrsquo Authentication and Copyright Protectionrdquo IEEETransactions on Circuits and Systems for Video Technology vol28 no 5 pp 1047ndash1055 2018
[29] Poonam and S M Arora ldquoA DWT-SVD based Robust DigitalWatermarking for Digital Imagesrdquo Procedia Computer Sciencevol 132 pp 1441ndash1448 2018
[30] S Fazli and M Moeini ldquoA robust image watermarking methodbased on DWT DCT and SVD using a new technique forcorrection of main geometric attacksrdquo Optik - InternationalJournal for Light and ElectronOptics vol 127 no 2 pp 964ndash9722016
[31] K Kuppusamy and K Thamodaran ldquoOptimized Image Water-marking Scheme Based On PSOrdquo Procedia Engineering vol 38pp 493ndash503 2012
[32] V S Rao R S Shekhawat and V K Srivastava ldquoA dwt-dct-svd based digital image watermarking scheme using particleswarm optimizationrdquo in Proceedings of the 2012 IEEE StudentsrsquoConference on Electrical Electronics and Computer Science pp1ndash4 2012
[33] T T Takore P R Kumar and G L Devi ldquoA modified blindimage watermarking scheme based on DWT DCT and SVDdomain using GA to optimize robustnessrdquo in Proceedings ofthe 2016 International Conference on Electrical Electronics andOptimization Techniques ICEEOT 2016 pp 2725ndash2729 IndiaMarch 2016
[34] H Hu S Lin and L Hsu ldquoEffective blind speech watermarkingvia adaptive mean modulation and package synchronization inDWT domainrdquo EURASIP Journal on Audio Speech and MusicProcessing vol 2017 no 1 2017
[35] W-H Lin Y-R Wang S-J Horng T-W Kao and Y Pan ldquoAblind watermarking method using maximum wavelet coeffi-cient quantizationrdquo Expert Systems with Applications vol 36no 9 pp 11509ndash11516 2009
[36] X Li J Liu J Sun X Yang andW Liu ldquoStep-projection-basedspread transform dither modulationrdquo IET Information Securityvol 5 no 3 pp 170ndash180 2011
[37] M Hamghalam S Mirzakuchaki andM A Akhaee ldquoGeomet-ric modelling of the wavelet coefficients for image watermark-ing using optimum detectorrdquo IET Image Processing vol 8 no 3pp 162ndash172 2014
[38] V S Verma R K Jha and A Ojha ldquoDigital watermarkextraction using support vector machine with principal com-ponent analysis based feature reductionrdquo Journal of VisualCommunication and Image Representation vol 31 pp 75ndash852015
[39] C Li Z Zhang Y Wang B Ma and D Huang ldquoDithermodulation of significant amplitude difference for waveletbased robust watermarkingrdquo Neurocomputing vol 166 pp404ndash415 2015
Figure 7 Computational time of (a) overall insertion and detection and (b) details of the embedding stage
restrict or expand the use of a given watermarking schemeIn the literature most of schemes are evaluated in a standardimage with a size of 512 times 512 pixels and there is noperspective of the computational cost trend when the imagegoes bigger which is more realistic Hence we propose tovary the size of the image beginning with 64 times 64 pixelsand ending with 4096 times 4096 pixels by doubling the heightand width after each insertion For the sake of simplicity thewatermark was embedded after one level of DWT (119897 = 1)with a fixed block size (119905 = 4) and a maximum watermarksize as defined by (8)
Figure 7(a) depicts the execution time evolution ofthe overall embedding and detection alongside with thevariation of the watermark size (red) as a function of theimage size The details of the embedding phase are shownin Figure 7(b) The most important remark to be made fromFigure 7 is that the computational time evolves broadly inline with the watermark size (red curve) to find out thecomplexity of our scheme in comparison with other state-ofthe-art methods
44 Robustness In the following we investigate the robust-ness of the suggested watermarking approach under severaltypes of attacks The robustness is quantified using the BitError Rate (BER) and the Normalized Correlation (NC)metrics
441 Evaluated Attacks Since the proposed scheme isdesigned to be robust the embedded message should lastor at list a significant part of it after image processingoperations Fromwatermark perspective all image processingoperations whatever performed on goodwill or not is consid-ered as attacks The involved attacks in the evaluation tests areclassified into three groups
(i) Compression attacks usually images are not storedin a raw format but after a compression operationthat aims to reduce the required storage space JPEGalgorithm is one of the most used compression
algorithms However JPEG is a lossy algorithm thatcauses some quality degradation inversely commen-surate with the compression quality factor Hence wetested the robustness of our algorithm under differentJPEG quality factors ranging from 10 to 90
(ii) Image processing attacks used to imitate a noisecontaminating the image during transmission such assalt and peppers or to improve the imagersquos qualityby eliminating a noise or enhancing some propertiesThe used attacks are listed as follows scaling medianfiltering sharpening Gaussian filtering average fil-tering Gaussian noise salt and pepper noise contam-ination histogram equalization and additive whiteGaussian noise (AWGN)
(iii) Geometric attacks used to correct geometric distor-tions such camera orientation (rotation) or imagesrsquoalignment (cropping) The used attacks are listed asfollows image cropping and resizing and rotation
Since the DWT coefficients are not geometrically invari-ant a manual synchronization of the imagersquos dimension andorientation was realized prior to the extraction
442 Performance Evaluation Here we test the robustnessof our scheme on three different images Boat Peppersand Barbara Each time we varied the embedded messagebetween 128 bits and 2048 bits The embedding threshold 119879is manually tuned to provide a watermarked image with aPSNR value of 40 dB and 45 dB The robustness is expressedin terms of BER () after mean filter (Figure 8) JPEGcompression (Figure 9) median filter (Figure 10) and gammacorrection Poisson noise histogram equalization and imageunsharpening (Figure 11) Figures 8ndash11 show that the robust-ness performance slightly differs between images For a smallmessage length (between 128 and 256 bits) and a PSNR valueof 40 dB the BER does not exceed 20 for most of attacksHowever the larger the message size is or the stronger theattack is the higher the BER will be
10 Security and Communication Networks
128 256 512 1024 2048Watermark size
0
10
20
30
40
50BE
R (
)
Mean filter (3x3)Mean filter (5x5)Mean filter (7x7)Mean filter (9x9)BarbaraPeppersBoat
(a)
Mean filter (3x3)Mean filter (5x5)Mean filter (7x7)Mean filter (9x9)BarbaraPeppersBoat
128 256 512 1024 2048Watermark size
0
10
20
30
40
50
BER
()
(b)
Figure 8 BER() after mean filter attacks using different message lengths at different PSNR values (a) 40 dB and (b) 45 dB
Figure 9 BER() after JPEG compression using different message lengths at different PSNR values (a) 40 dB and (b) 45 dB
443 Comparative Study To prove the superiority and therobustness advantages of our scheme a comparative studywith state of the art works is indispensable In fact wecompared the robustness of our scheme with six robustwatermarking schemes [26 27 35 37ndash39] In order to carryout a fair comparison the same insertion parameters (PSNRwatermark size) and attacks are used in this comparisonThe PSNRs value of different images and their correspondingvalues in the comparative schemes are illustrated in Figure 12It is worth mentioning that the insertion parameters (119897 119905 119879)were empirically chosen to produce the same PSNR alongsidewith an acceptable visual quality
The robustness results are presented in Tables 2ndash5 interms of either BER or NC depending on how they appearin the reference papers and the superior results are markedin bold number
Based on the simulation results given in Tables 2ndash5 it canbe found that our approach outperforms similar schemes on7194 of the time Equal results are obtained on 2343 ofthe time and our scheme underperforms on 462 of thecases
45 Security The robustness of a watermarking scheme isnot sufficient to judge a scheme as reliable As discussedin the first section robust watermarking is used to secureimages Hence the security is necessary First the watermarkshould be immune against illegal extraction even when theembedding procedure is exposed Second the watermarkdetector should not report the existence of the watermarkwhen none is embedded To evaluate these two risks weextract the watermark from a watermarked image using
Security and Communication Networks 11
Table 2 BER comparison with [27] and [37] using a message length of 128 bits
different keys and extract a watermark from a nonwater-marked image respectively In the two cases the maximumcorrelation obtained has not exceeded the value 059 whichpoints out a negative presence
5 Conclusion
The mean value of DWTrsquos vertical and horizontal subbandscoefficients is highly robust and imperceptible In fact these
two advantages were exploited in the design of a robustand blind watermarking approach The embedding methodconsisted in modulating the BMV in accordance with thequantization technique The use of a single transformationdomain alongside with a simple embedding technique makesfrom the proposed algorithmcomputationally undemandingAt the extraction phase the watermark bits were identifiedby the BMV position with respect to the zero value Unlikesome existing schemes the extraction rules did not require to
Security and Communication Networks 15
be adjusted to the applied attacks to fulfill a high robustnessIn fact the inserted watermark was capable to withstand withno error after JPEG compression up to a quality factor of 30Gaussian filter and other attacks
Data Availability
The data used to support the findings of this study areincluded within the article
Conflicts of Interest
The authors declare that they have no conflicts of interest
References
[1] S Liansheng Z Bei W Zhanmin and T Ailing ldquoAn opticalcolor image watermarking scheme by using compressive sens-ingwith human visual characteristics in gyrator domainrdquoOpticsand Lasers in Engineering vol 92 pp 85ndash93 2017
[2] D Shukla and M Sharma ldquoA Novel Scene-Based Video Water-marking Scheme for Copyright Protectionrdquo Journal of IntelligentSystems vol 27 no 1 pp 47ndash66 2018
[3] A Tiwari and M Sharma ldquoAn efficient vector quantizationbased watermarking method for image integrity authentica-tionrdquo Advances in Intelligent Systems and Computing vol 519pp 215ndash225 2018
[4] S Bravo-Solorio F Calderon C-T Li and A K Nandi ldquoFastfragile watermark embedding and iterative mechanism withhigh self-restoration performancerdquo Digital Signal Processingvol 73 pp 83ndash92 2018
[5] A Ghardallou A Kricha A Sakly and A Mtibaa ldquoAdaptiveblock sized reversible watermarking scheme based on integertransformrdquo in Proceedings of the 17th International Conferenceon Sciences and Techniques of Automatic Control and ComputerEngineering STA 2016 pp 347ndash351 Tunisia December 2016
[6] R G van Schyndel A Z Tirkel and C F Osborne ldquoA digitalwatermarkrdquo in Proceedings of the IEEE International ConferenceImage Processing (ICIP rsquo94) vol 2 pp 86ndash90 Austin Tex USANovember 1994
[7] W-H Lin S-J Horng T-W Kao P Fan C-L Lee and Y PanldquoAn efficient watermarking method based on significant differ-ence of wavelet coefficient quantizationrdquo IEEE Transactions onMultimedia vol 10 no 5 pp 746ndash757 2008
[8] S Bhattacharyya and G Sanyal ldquoA Robust Image Steganogra-phy using DWT Difference Modulation (DWTDM)rdquo Interna-tional Journal of Computer Network and Information Securityvol 4 no 7 pp 27ndash40 2012
[9] H S Malvar and D A Florencio ldquoImproved spread spectruma new modulation technique for robust watermarkingrdquo IEEETransactions on Signal Processing vol 51 no 4 pp 898ndash9052003
[10] B Chen and G W Wornell ldquoProvably robust digital water-markingrdquo in Proceedings of the 1999 Multimedia Systems andApplications II pp 43ndash54 September 1999
[11] Z Kricha A Kricha and A Sakly ldquoAccommodative extractorfor QIM-based watermarking schemesrdquo IET Image Processing2018
[12] I J Cox J Kilian T Leighton and T Shamoon ldquoSecure imper-ceptable yet perceptually salient spread spectrum watermark
formultimediardquo in Proceedings of the 1996 Southcon Conferencepp 192ndash197 June 1996
[13] X Li J Liu J Sun Z Sun H Hu and L Zhang ldquoImprovedspread transform dither modulation-based watermark algo-rithm based on step projectionrdquo in Proceedings of the 2010 IEEEInternational Symposium on Broadband Multimedia Systemsand Broadcasting (BMSB) pp 1ndash5 Shanghai China March2010
[14] H-T Hu and L-Y Hsu ldquoExploring DWT-SVD-DCT featureparameters for robust multiple watermarking against JPEG andJPEG2000 compressionrdquo Computers and Electrical Engineeringvol 41 pp 52ndash63 2015
[15] J Cannons and P Moulin ldquoDesign and statistical analysis of ahash-aided image watermarking systemrdquo IEEE Transactions onImage Processing vol 13 no 10 pp 1393ndash1408 2004
[16] A Valizadeh and Z J Wang ldquoAn improved multiplicativespread spectrum embedding scheme for data hidingrdquo IEEETransactions on Information Forensics and Security vol 7 no4 pp 1127ndash1143 2012
[17] M Kutter D Frederic and F B JordanDigital signature of colorimages using amplitude modulation 1997
[18] Q Su and B Chen ldquoRobust color image watermarking tech-nique in the spatial domainrdquo Soft Computing vol 22 no 1 pp91ndash106 2018
[19] Z Yanling Z Xiaoshi L Na et al ldquoA Digital ImageWatermarkAlgorithm Based onDCCoefficientsQuantizationrdquo in Proceed-ings of the 2006 6th World Congress on Intelligent Control andAutomation pp 9734ndash9738 Dalian China 2006
[20] RMZhaoH LianHWPang andBNHu ldquoAWatermarkingAlgorithm by Modifying AC Coefficies in DCT Domainrdquo inProceedings of the 2008 International Symposium on InformationScience and Engineering (ISISE) pp 159ndash162 Shanghai ChinaDecember 2008
[21] B L Gunjal and R RManthalkar ldquoDiscreteWavelet Transformbased strongly robust watermarking scheme for informationhiding in digital imagesrdquo in Proceedings of the 3rd InternationalConference on Emerging Trends in Engineering and TechnologyICETET 2010 pp 124ndash129 India November 2010
[22] H Shi and F Lv ldquoA Blind Digital Watermark Techniquefor Color Image Based on Integer Wavelet Transformrdquo inProceedings of the 2010 International Conference on BiomedicalEngineering and Computer Science (ICBECS) pp 1ndash4 WuhanChina April 2010
[23] S Ranjbar F Zargari and M Ghanbari ldquoA highly robust two-stage Contourlet-based digital image watermarking methodrdquoSignal Processing Image Communication vol 28 no 10 pp1526ndash1536 2013
[24] Q Su and B Chen ldquoA novel blind color image watermarkingusing upper Hessenbergmatrixrdquo AEU - International Journal ofElectronics and Communications vol 78 pp 64ndash71 2017
[25] F Ourique V Licks R Jordan and F Perez-Gonzalez ldquoAngleQIM a novel watermark embedding scheme robust againstamplitude scaling distortionsrdquo in Proceedings of the IEEE Inter-national Conference on Acoustics Speech and Signal Processing(ICASSP rsquo05) pp 797ndash800 March 2005
[26] H R Fazlali S Samavi N Karimi and S Shirani ldquoAdaptiveblind image watermarking using edge pixel concentrationrdquoMultimedia Tools and Applications vol 76 no 2 pp 3105ndash31202017
[27] X-Y Wang Y-N Liu H Xu A-L Wang and H-Y YangldquoBlind optimum detector for robust image watermarking in
16 Security and Communication Networks
nonsubsampled shearlet Domainrdquo Information Sciences vol372 pp 634ndash654 2016
[28] X Liu C Lin and S Yuan ldquoBlind Dual Watermarking forColor Imagesrsquo Authentication and Copyright Protectionrdquo IEEETransactions on Circuits and Systems for Video Technology vol28 no 5 pp 1047ndash1055 2018
[29] Poonam and S M Arora ldquoA DWT-SVD based Robust DigitalWatermarking for Digital Imagesrdquo Procedia Computer Sciencevol 132 pp 1441ndash1448 2018
[30] S Fazli and M Moeini ldquoA robust image watermarking methodbased on DWT DCT and SVD using a new technique forcorrection of main geometric attacksrdquo Optik - InternationalJournal for Light and ElectronOptics vol 127 no 2 pp 964ndash9722016
[31] K Kuppusamy and K Thamodaran ldquoOptimized Image Water-marking Scheme Based On PSOrdquo Procedia Engineering vol 38pp 493ndash503 2012
[32] V S Rao R S Shekhawat and V K Srivastava ldquoA dwt-dct-svd based digital image watermarking scheme using particleswarm optimizationrdquo in Proceedings of the 2012 IEEE StudentsrsquoConference on Electrical Electronics and Computer Science pp1ndash4 2012
[33] T T Takore P R Kumar and G L Devi ldquoA modified blindimage watermarking scheme based on DWT DCT and SVDdomain using GA to optimize robustnessrdquo in Proceedings ofthe 2016 International Conference on Electrical Electronics andOptimization Techniques ICEEOT 2016 pp 2725ndash2729 IndiaMarch 2016
[34] H Hu S Lin and L Hsu ldquoEffective blind speech watermarkingvia adaptive mean modulation and package synchronization inDWT domainrdquo EURASIP Journal on Audio Speech and MusicProcessing vol 2017 no 1 2017
[35] W-H Lin Y-R Wang S-J Horng T-W Kao and Y Pan ldquoAblind watermarking method using maximum wavelet coeffi-cient quantizationrdquo Expert Systems with Applications vol 36no 9 pp 11509ndash11516 2009
[36] X Li J Liu J Sun X Yang andW Liu ldquoStep-projection-basedspread transform dither modulationrdquo IET Information Securityvol 5 no 3 pp 170ndash180 2011
[37] M Hamghalam S Mirzakuchaki andM A Akhaee ldquoGeomet-ric modelling of the wavelet coefficients for image watermark-ing using optimum detectorrdquo IET Image Processing vol 8 no 3pp 162ndash172 2014
[38] V S Verma R K Jha and A Ojha ldquoDigital watermarkextraction using support vector machine with principal com-ponent analysis based feature reductionrdquo Journal of VisualCommunication and Image Representation vol 31 pp 75ndash852015
[39] C Li Z Zhang Y Wang B Ma and D Huang ldquoDithermodulation of significant amplitude difference for waveletbased robust watermarkingrdquo Neurocomputing vol 166 pp404ndash415 2015
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Navigation and Observation
International Journal of
Hindawi
wwwhindawicom Volume 2018
Advances in
Multimedia
Submit your manuscripts atwwwhindawicom
Security and Communication Networks 9
05
1015202530354045
64 128
256
512
1024
2048
4096
Com
puta
tion
time (
s)
Image heightwidth (pixels)
Message size (bits)
25
2
15
1
05
0
InsertionDetectionMessage size
106times
(a)
Image heightwidth (pixels)
Message size (bits)
05
1015202530354045
64 128
256
512
1024
2048
4096
Com
puta
tion
time (
s)
25
2
15
1
05
0
DWTiDWTChaotic sequence
Mean ModulationMessage size
106times
(b)
Figure 7 Computational time of (a) overall insertion and detection and (b) details of the embedding stage
restrict or expand the use of a given watermarking schemeIn the literature most of schemes are evaluated in a standardimage with a size of 512 times 512 pixels and there is noperspective of the computational cost trend when the imagegoes bigger which is more realistic Hence we propose tovary the size of the image beginning with 64 times 64 pixelsand ending with 4096 times 4096 pixels by doubling the heightand width after each insertion For the sake of simplicity thewatermark was embedded after one level of DWT (119897 = 1)with a fixed block size (119905 = 4) and a maximum watermarksize as defined by (8)
Figure 7(a) depicts the execution time evolution ofthe overall embedding and detection alongside with thevariation of the watermark size (red) as a function of theimage size The details of the embedding phase are shownin Figure 7(b) The most important remark to be made fromFigure 7 is that the computational time evolves broadly inline with the watermark size (red curve) to find out thecomplexity of our scheme in comparison with other state-ofthe-art methods
44 Robustness In the following we investigate the robust-ness of the suggested watermarking approach under severaltypes of attacks The robustness is quantified using the BitError Rate (BER) and the Normalized Correlation (NC)metrics
441 Evaluated Attacks Since the proposed scheme isdesigned to be robust the embedded message should lastor at list a significant part of it after image processingoperations Fromwatermark perspective all image processingoperations whatever performed on goodwill or not is consid-ered as attacks The involved attacks in the evaluation tests areclassified into three groups
(i) Compression attacks usually images are not storedin a raw format but after a compression operationthat aims to reduce the required storage space JPEGalgorithm is one of the most used compression
algorithms However JPEG is a lossy algorithm thatcauses some quality degradation inversely commen-surate with the compression quality factor Hence wetested the robustness of our algorithm under differentJPEG quality factors ranging from 10 to 90
(ii) Image processing attacks used to imitate a noisecontaminating the image during transmission such assalt and peppers or to improve the imagersquos qualityby eliminating a noise or enhancing some propertiesThe used attacks are listed as follows scaling medianfiltering sharpening Gaussian filtering average fil-tering Gaussian noise salt and pepper noise contam-ination histogram equalization and additive whiteGaussian noise (AWGN)
(iii) Geometric attacks used to correct geometric distor-tions such camera orientation (rotation) or imagesrsquoalignment (cropping) The used attacks are listed asfollows image cropping and resizing and rotation
Since the DWT coefficients are not geometrically invari-ant a manual synchronization of the imagersquos dimension andorientation was realized prior to the extraction
442 Performance Evaluation Here we test the robustnessof our scheme on three different images Boat Peppersand Barbara Each time we varied the embedded messagebetween 128 bits and 2048 bits The embedding threshold 119879is manually tuned to provide a watermarked image with aPSNR value of 40 dB and 45 dB The robustness is expressedin terms of BER () after mean filter (Figure 8) JPEGcompression (Figure 9) median filter (Figure 10) and gammacorrection Poisson noise histogram equalization and imageunsharpening (Figure 11) Figures 8ndash11 show that the robust-ness performance slightly differs between images For a smallmessage length (between 128 and 256 bits) and a PSNR valueof 40 dB the BER does not exceed 20 for most of attacksHowever the larger the message size is or the stronger theattack is the higher the BER will be
10 Security and Communication Networks
128 256 512 1024 2048Watermark size
0
10
20
30
40
50BE
R (
)
Mean filter (3x3)Mean filter (5x5)Mean filter (7x7)Mean filter (9x9)BarbaraPeppersBoat
(a)
Mean filter (3x3)Mean filter (5x5)Mean filter (7x7)Mean filter (9x9)BarbaraPeppersBoat
128 256 512 1024 2048Watermark size
0
10
20
30
40
50
BER
()
(b)
Figure 8 BER() after mean filter attacks using different message lengths at different PSNR values (a) 40 dB and (b) 45 dB
Figure 9 BER() after JPEG compression using different message lengths at different PSNR values (a) 40 dB and (b) 45 dB
443 Comparative Study To prove the superiority and therobustness advantages of our scheme a comparative studywith state of the art works is indispensable In fact wecompared the robustness of our scheme with six robustwatermarking schemes [26 27 35 37ndash39] In order to carryout a fair comparison the same insertion parameters (PSNRwatermark size) and attacks are used in this comparisonThe PSNRs value of different images and their correspondingvalues in the comparative schemes are illustrated in Figure 12It is worth mentioning that the insertion parameters (119897 119905 119879)were empirically chosen to produce the same PSNR alongsidewith an acceptable visual quality
The robustness results are presented in Tables 2ndash5 interms of either BER or NC depending on how they appearin the reference papers and the superior results are markedin bold number
Based on the simulation results given in Tables 2ndash5 it canbe found that our approach outperforms similar schemes on7194 of the time Equal results are obtained on 2343 ofthe time and our scheme underperforms on 462 of thecases
45 Security The robustness of a watermarking scheme isnot sufficient to judge a scheme as reliable As discussedin the first section robust watermarking is used to secureimages Hence the security is necessary First the watermarkshould be immune against illegal extraction even when theembedding procedure is exposed Second the watermarkdetector should not report the existence of the watermarkwhen none is embedded To evaluate these two risks weextract the watermark from a watermarked image using
Security and Communication Networks 11
Table 2 BER comparison with [27] and [37] using a message length of 128 bits
different keys and extract a watermark from a nonwater-marked image respectively In the two cases the maximumcorrelation obtained has not exceeded the value 059 whichpoints out a negative presence
5 Conclusion
The mean value of DWTrsquos vertical and horizontal subbandscoefficients is highly robust and imperceptible In fact these
two advantages were exploited in the design of a robustand blind watermarking approach The embedding methodconsisted in modulating the BMV in accordance with thequantization technique The use of a single transformationdomain alongside with a simple embedding technique makesfrom the proposed algorithmcomputationally undemandingAt the extraction phase the watermark bits were identifiedby the BMV position with respect to the zero value Unlikesome existing schemes the extraction rules did not require to
Security and Communication Networks 15
be adjusted to the applied attacks to fulfill a high robustnessIn fact the inserted watermark was capable to withstand withno error after JPEG compression up to a quality factor of 30Gaussian filter and other attacks
Data Availability
The data used to support the findings of this study areincluded within the article
Conflicts of Interest
The authors declare that they have no conflicts of interest
References
[1] S Liansheng Z Bei W Zhanmin and T Ailing ldquoAn opticalcolor image watermarking scheme by using compressive sens-ingwith human visual characteristics in gyrator domainrdquoOpticsand Lasers in Engineering vol 92 pp 85ndash93 2017
[2] D Shukla and M Sharma ldquoA Novel Scene-Based Video Water-marking Scheme for Copyright Protectionrdquo Journal of IntelligentSystems vol 27 no 1 pp 47ndash66 2018
[3] A Tiwari and M Sharma ldquoAn efficient vector quantizationbased watermarking method for image integrity authentica-tionrdquo Advances in Intelligent Systems and Computing vol 519pp 215ndash225 2018
[4] S Bravo-Solorio F Calderon C-T Li and A K Nandi ldquoFastfragile watermark embedding and iterative mechanism withhigh self-restoration performancerdquo Digital Signal Processingvol 73 pp 83ndash92 2018
[5] A Ghardallou A Kricha A Sakly and A Mtibaa ldquoAdaptiveblock sized reversible watermarking scheme based on integertransformrdquo in Proceedings of the 17th International Conferenceon Sciences and Techniques of Automatic Control and ComputerEngineering STA 2016 pp 347ndash351 Tunisia December 2016
[6] R G van Schyndel A Z Tirkel and C F Osborne ldquoA digitalwatermarkrdquo in Proceedings of the IEEE International ConferenceImage Processing (ICIP rsquo94) vol 2 pp 86ndash90 Austin Tex USANovember 1994
[7] W-H Lin S-J Horng T-W Kao P Fan C-L Lee and Y PanldquoAn efficient watermarking method based on significant differ-ence of wavelet coefficient quantizationrdquo IEEE Transactions onMultimedia vol 10 no 5 pp 746ndash757 2008
[8] S Bhattacharyya and G Sanyal ldquoA Robust Image Steganogra-phy using DWT Difference Modulation (DWTDM)rdquo Interna-tional Journal of Computer Network and Information Securityvol 4 no 7 pp 27ndash40 2012
[9] H S Malvar and D A Florencio ldquoImproved spread spectruma new modulation technique for robust watermarkingrdquo IEEETransactions on Signal Processing vol 51 no 4 pp 898ndash9052003
[10] B Chen and G W Wornell ldquoProvably robust digital water-markingrdquo in Proceedings of the 1999 Multimedia Systems andApplications II pp 43ndash54 September 1999
[11] Z Kricha A Kricha and A Sakly ldquoAccommodative extractorfor QIM-based watermarking schemesrdquo IET Image Processing2018
[12] I J Cox J Kilian T Leighton and T Shamoon ldquoSecure imper-ceptable yet perceptually salient spread spectrum watermark
formultimediardquo in Proceedings of the 1996 Southcon Conferencepp 192ndash197 June 1996
[13] X Li J Liu J Sun Z Sun H Hu and L Zhang ldquoImprovedspread transform dither modulation-based watermark algo-rithm based on step projectionrdquo in Proceedings of the 2010 IEEEInternational Symposium on Broadband Multimedia Systemsand Broadcasting (BMSB) pp 1ndash5 Shanghai China March2010
[14] H-T Hu and L-Y Hsu ldquoExploring DWT-SVD-DCT featureparameters for robust multiple watermarking against JPEG andJPEG2000 compressionrdquo Computers and Electrical Engineeringvol 41 pp 52ndash63 2015
[15] J Cannons and P Moulin ldquoDesign and statistical analysis of ahash-aided image watermarking systemrdquo IEEE Transactions onImage Processing vol 13 no 10 pp 1393ndash1408 2004
[16] A Valizadeh and Z J Wang ldquoAn improved multiplicativespread spectrum embedding scheme for data hidingrdquo IEEETransactions on Information Forensics and Security vol 7 no4 pp 1127ndash1143 2012
[17] M Kutter D Frederic and F B JordanDigital signature of colorimages using amplitude modulation 1997
[18] Q Su and B Chen ldquoRobust color image watermarking tech-nique in the spatial domainrdquo Soft Computing vol 22 no 1 pp91ndash106 2018
[19] Z Yanling Z Xiaoshi L Na et al ldquoA Digital ImageWatermarkAlgorithm Based onDCCoefficientsQuantizationrdquo in Proceed-ings of the 2006 6th World Congress on Intelligent Control andAutomation pp 9734ndash9738 Dalian China 2006
[20] RMZhaoH LianHWPang andBNHu ldquoAWatermarkingAlgorithm by Modifying AC Coefficies in DCT Domainrdquo inProceedings of the 2008 International Symposium on InformationScience and Engineering (ISISE) pp 159ndash162 Shanghai ChinaDecember 2008
[21] B L Gunjal and R RManthalkar ldquoDiscreteWavelet Transformbased strongly robust watermarking scheme for informationhiding in digital imagesrdquo in Proceedings of the 3rd InternationalConference on Emerging Trends in Engineering and TechnologyICETET 2010 pp 124ndash129 India November 2010
[22] H Shi and F Lv ldquoA Blind Digital Watermark Techniquefor Color Image Based on Integer Wavelet Transformrdquo inProceedings of the 2010 International Conference on BiomedicalEngineering and Computer Science (ICBECS) pp 1ndash4 WuhanChina April 2010
[23] S Ranjbar F Zargari and M Ghanbari ldquoA highly robust two-stage Contourlet-based digital image watermarking methodrdquoSignal Processing Image Communication vol 28 no 10 pp1526ndash1536 2013
[24] Q Su and B Chen ldquoA novel blind color image watermarkingusing upper Hessenbergmatrixrdquo AEU - International Journal ofElectronics and Communications vol 78 pp 64ndash71 2017
[25] F Ourique V Licks R Jordan and F Perez-Gonzalez ldquoAngleQIM a novel watermark embedding scheme robust againstamplitude scaling distortionsrdquo in Proceedings of the IEEE Inter-national Conference on Acoustics Speech and Signal Processing(ICASSP rsquo05) pp 797ndash800 March 2005
[26] H R Fazlali S Samavi N Karimi and S Shirani ldquoAdaptiveblind image watermarking using edge pixel concentrationrdquoMultimedia Tools and Applications vol 76 no 2 pp 3105ndash31202017
[27] X-Y Wang Y-N Liu H Xu A-L Wang and H-Y YangldquoBlind optimum detector for robust image watermarking in
16 Security and Communication Networks
nonsubsampled shearlet Domainrdquo Information Sciences vol372 pp 634ndash654 2016
[28] X Liu C Lin and S Yuan ldquoBlind Dual Watermarking forColor Imagesrsquo Authentication and Copyright Protectionrdquo IEEETransactions on Circuits and Systems for Video Technology vol28 no 5 pp 1047ndash1055 2018
[29] Poonam and S M Arora ldquoA DWT-SVD based Robust DigitalWatermarking for Digital Imagesrdquo Procedia Computer Sciencevol 132 pp 1441ndash1448 2018
[30] S Fazli and M Moeini ldquoA robust image watermarking methodbased on DWT DCT and SVD using a new technique forcorrection of main geometric attacksrdquo Optik - InternationalJournal for Light and ElectronOptics vol 127 no 2 pp 964ndash9722016
[31] K Kuppusamy and K Thamodaran ldquoOptimized Image Water-marking Scheme Based On PSOrdquo Procedia Engineering vol 38pp 493ndash503 2012
[32] V S Rao R S Shekhawat and V K Srivastava ldquoA dwt-dct-svd based digital image watermarking scheme using particleswarm optimizationrdquo in Proceedings of the 2012 IEEE StudentsrsquoConference on Electrical Electronics and Computer Science pp1ndash4 2012
[33] T T Takore P R Kumar and G L Devi ldquoA modified blindimage watermarking scheme based on DWT DCT and SVDdomain using GA to optimize robustnessrdquo in Proceedings ofthe 2016 International Conference on Electrical Electronics andOptimization Techniques ICEEOT 2016 pp 2725ndash2729 IndiaMarch 2016
[34] H Hu S Lin and L Hsu ldquoEffective blind speech watermarkingvia adaptive mean modulation and package synchronization inDWT domainrdquo EURASIP Journal on Audio Speech and MusicProcessing vol 2017 no 1 2017
[35] W-H Lin Y-R Wang S-J Horng T-W Kao and Y Pan ldquoAblind watermarking method using maximum wavelet coeffi-cient quantizationrdquo Expert Systems with Applications vol 36no 9 pp 11509ndash11516 2009
[36] X Li J Liu J Sun X Yang andW Liu ldquoStep-projection-basedspread transform dither modulationrdquo IET Information Securityvol 5 no 3 pp 170ndash180 2011
[37] M Hamghalam S Mirzakuchaki andM A Akhaee ldquoGeomet-ric modelling of the wavelet coefficients for image watermark-ing using optimum detectorrdquo IET Image Processing vol 8 no 3pp 162ndash172 2014
[38] V S Verma R K Jha and A Ojha ldquoDigital watermarkextraction using support vector machine with principal com-ponent analysis based feature reductionrdquo Journal of VisualCommunication and Image Representation vol 31 pp 75ndash852015
[39] C Li Z Zhang Y Wang B Ma and D Huang ldquoDithermodulation of significant amplitude difference for waveletbased robust watermarkingrdquo Neurocomputing vol 166 pp404ndash415 2015
Figure 9 BER() after JPEG compression using different message lengths at different PSNR values (a) 40 dB and (b) 45 dB
443 Comparative Study To prove the superiority and therobustness advantages of our scheme a comparative studywith state of the art works is indispensable In fact wecompared the robustness of our scheme with six robustwatermarking schemes [26 27 35 37ndash39] In order to carryout a fair comparison the same insertion parameters (PSNRwatermark size) and attacks are used in this comparisonThe PSNRs value of different images and their correspondingvalues in the comparative schemes are illustrated in Figure 12It is worth mentioning that the insertion parameters (119897 119905 119879)were empirically chosen to produce the same PSNR alongsidewith an acceptable visual quality
The robustness results are presented in Tables 2ndash5 interms of either BER or NC depending on how they appearin the reference papers and the superior results are markedin bold number
Based on the simulation results given in Tables 2ndash5 it canbe found that our approach outperforms similar schemes on7194 of the time Equal results are obtained on 2343 ofthe time and our scheme underperforms on 462 of thecases
45 Security The robustness of a watermarking scheme isnot sufficient to judge a scheme as reliable As discussedin the first section robust watermarking is used to secureimages Hence the security is necessary First the watermarkshould be immune against illegal extraction even when theembedding procedure is exposed Second the watermarkdetector should not report the existence of the watermarkwhen none is embedded To evaluate these two risks weextract the watermark from a watermarked image using
Security and Communication Networks 11
Table 2 BER comparison with [27] and [37] using a message length of 128 bits
different keys and extract a watermark from a nonwater-marked image respectively In the two cases the maximumcorrelation obtained has not exceeded the value 059 whichpoints out a negative presence
5 Conclusion
The mean value of DWTrsquos vertical and horizontal subbandscoefficients is highly robust and imperceptible In fact these
two advantages were exploited in the design of a robustand blind watermarking approach The embedding methodconsisted in modulating the BMV in accordance with thequantization technique The use of a single transformationdomain alongside with a simple embedding technique makesfrom the proposed algorithmcomputationally undemandingAt the extraction phase the watermark bits were identifiedby the BMV position with respect to the zero value Unlikesome existing schemes the extraction rules did not require to
Security and Communication Networks 15
be adjusted to the applied attacks to fulfill a high robustnessIn fact the inserted watermark was capable to withstand withno error after JPEG compression up to a quality factor of 30Gaussian filter and other attacks
Data Availability
The data used to support the findings of this study areincluded within the article
Conflicts of Interest
The authors declare that they have no conflicts of interest
References
[1] S Liansheng Z Bei W Zhanmin and T Ailing ldquoAn opticalcolor image watermarking scheme by using compressive sens-ingwith human visual characteristics in gyrator domainrdquoOpticsand Lasers in Engineering vol 92 pp 85ndash93 2017
[2] D Shukla and M Sharma ldquoA Novel Scene-Based Video Water-marking Scheme for Copyright Protectionrdquo Journal of IntelligentSystems vol 27 no 1 pp 47ndash66 2018
[3] A Tiwari and M Sharma ldquoAn efficient vector quantizationbased watermarking method for image integrity authentica-tionrdquo Advances in Intelligent Systems and Computing vol 519pp 215ndash225 2018
[4] S Bravo-Solorio F Calderon C-T Li and A K Nandi ldquoFastfragile watermark embedding and iterative mechanism withhigh self-restoration performancerdquo Digital Signal Processingvol 73 pp 83ndash92 2018
[5] A Ghardallou A Kricha A Sakly and A Mtibaa ldquoAdaptiveblock sized reversible watermarking scheme based on integertransformrdquo in Proceedings of the 17th International Conferenceon Sciences and Techniques of Automatic Control and ComputerEngineering STA 2016 pp 347ndash351 Tunisia December 2016
[6] R G van Schyndel A Z Tirkel and C F Osborne ldquoA digitalwatermarkrdquo in Proceedings of the IEEE International ConferenceImage Processing (ICIP rsquo94) vol 2 pp 86ndash90 Austin Tex USANovember 1994
[7] W-H Lin S-J Horng T-W Kao P Fan C-L Lee and Y PanldquoAn efficient watermarking method based on significant differ-ence of wavelet coefficient quantizationrdquo IEEE Transactions onMultimedia vol 10 no 5 pp 746ndash757 2008
[8] S Bhattacharyya and G Sanyal ldquoA Robust Image Steganogra-phy using DWT Difference Modulation (DWTDM)rdquo Interna-tional Journal of Computer Network and Information Securityvol 4 no 7 pp 27ndash40 2012
[9] H S Malvar and D A Florencio ldquoImproved spread spectruma new modulation technique for robust watermarkingrdquo IEEETransactions on Signal Processing vol 51 no 4 pp 898ndash9052003
[10] B Chen and G W Wornell ldquoProvably robust digital water-markingrdquo in Proceedings of the 1999 Multimedia Systems andApplications II pp 43ndash54 September 1999
[11] Z Kricha A Kricha and A Sakly ldquoAccommodative extractorfor QIM-based watermarking schemesrdquo IET Image Processing2018
[12] I J Cox J Kilian T Leighton and T Shamoon ldquoSecure imper-ceptable yet perceptually salient spread spectrum watermark
formultimediardquo in Proceedings of the 1996 Southcon Conferencepp 192ndash197 June 1996
[13] X Li J Liu J Sun Z Sun H Hu and L Zhang ldquoImprovedspread transform dither modulation-based watermark algo-rithm based on step projectionrdquo in Proceedings of the 2010 IEEEInternational Symposium on Broadband Multimedia Systemsand Broadcasting (BMSB) pp 1ndash5 Shanghai China March2010
[14] H-T Hu and L-Y Hsu ldquoExploring DWT-SVD-DCT featureparameters for robust multiple watermarking against JPEG andJPEG2000 compressionrdquo Computers and Electrical Engineeringvol 41 pp 52ndash63 2015
[15] J Cannons and P Moulin ldquoDesign and statistical analysis of ahash-aided image watermarking systemrdquo IEEE Transactions onImage Processing vol 13 no 10 pp 1393ndash1408 2004
[16] A Valizadeh and Z J Wang ldquoAn improved multiplicativespread spectrum embedding scheme for data hidingrdquo IEEETransactions on Information Forensics and Security vol 7 no4 pp 1127ndash1143 2012
[17] M Kutter D Frederic and F B JordanDigital signature of colorimages using amplitude modulation 1997
[18] Q Su and B Chen ldquoRobust color image watermarking tech-nique in the spatial domainrdquo Soft Computing vol 22 no 1 pp91ndash106 2018
[19] Z Yanling Z Xiaoshi L Na et al ldquoA Digital ImageWatermarkAlgorithm Based onDCCoefficientsQuantizationrdquo in Proceed-ings of the 2006 6th World Congress on Intelligent Control andAutomation pp 9734ndash9738 Dalian China 2006
[20] RMZhaoH LianHWPang andBNHu ldquoAWatermarkingAlgorithm by Modifying AC Coefficies in DCT Domainrdquo inProceedings of the 2008 International Symposium on InformationScience and Engineering (ISISE) pp 159ndash162 Shanghai ChinaDecember 2008
[21] B L Gunjal and R RManthalkar ldquoDiscreteWavelet Transformbased strongly robust watermarking scheme for informationhiding in digital imagesrdquo in Proceedings of the 3rd InternationalConference on Emerging Trends in Engineering and TechnologyICETET 2010 pp 124ndash129 India November 2010
[22] H Shi and F Lv ldquoA Blind Digital Watermark Techniquefor Color Image Based on Integer Wavelet Transformrdquo inProceedings of the 2010 International Conference on BiomedicalEngineering and Computer Science (ICBECS) pp 1ndash4 WuhanChina April 2010
[23] S Ranjbar F Zargari and M Ghanbari ldquoA highly robust two-stage Contourlet-based digital image watermarking methodrdquoSignal Processing Image Communication vol 28 no 10 pp1526ndash1536 2013
[24] Q Su and B Chen ldquoA novel blind color image watermarkingusing upper Hessenbergmatrixrdquo AEU - International Journal ofElectronics and Communications vol 78 pp 64ndash71 2017
[25] F Ourique V Licks R Jordan and F Perez-Gonzalez ldquoAngleQIM a novel watermark embedding scheme robust againstamplitude scaling distortionsrdquo in Proceedings of the IEEE Inter-national Conference on Acoustics Speech and Signal Processing(ICASSP rsquo05) pp 797ndash800 March 2005
[26] H R Fazlali S Samavi N Karimi and S Shirani ldquoAdaptiveblind image watermarking using edge pixel concentrationrdquoMultimedia Tools and Applications vol 76 no 2 pp 3105ndash31202017
[27] X-Y Wang Y-N Liu H Xu A-L Wang and H-Y YangldquoBlind optimum detector for robust image watermarking in
16 Security and Communication Networks
nonsubsampled shearlet Domainrdquo Information Sciences vol372 pp 634ndash654 2016
[28] X Liu C Lin and S Yuan ldquoBlind Dual Watermarking forColor Imagesrsquo Authentication and Copyright Protectionrdquo IEEETransactions on Circuits and Systems for Video Technology vol28 no 5 pp 1047ndash1055 2018
[29] Poonam and S M Arora ldquoA DWT-SVD based Robust DigitalWatermarking for Digital Imagesrdquo Procedia Computer Sciencevol 132 pp 1441ndash1448 2018
[30] S Fazli and M Moeini ldquoA robust image watermarking methodbased on DWT DCT and SVD using a new technique forcorrection of main geometric attacksrdquo Optik - InternationalJournal for Light and ElectronOptics vol 127 no 2 pp 964ndash9722016
[31] K Kuppusamy and K Thamodaran ldquoOptimized Image Water-marking Scheme Based On PSOrdquo Procedia Engineering vol 38pp 493ndash503 2012
[32] V S Rao R S Shekhawat and V K Srivastava ldquoA dwt-dct-svd based digital image watermarking scheme using particleswarm optimizationrdquo in Proceedings of the 2012 IEEE StudentsrsquoConference on Electrical Electronics and Computer Science pp1ndash4 2012
[33] T T Takore P R Kumar and G L Devi ldquoA modified blindimage watermarking scheme based on DWT DCT and SVDdomain using GA to optimize robustnessrdquo in Proceedings ofthe 2016 International Conference on Electrical Electronics andOptimization Techniques ICEEOT 2016 pp 2725ndash2729 IndiaMarch 2016
[34] H Hu S Lin and L Hsu ldquoEffective blind speech watermarkingvia adaptive mean modulation and package synchronization inDWT domainrdquo EURASIP Journal on Audio Speech and MusicProcessing vol 2017 no 1 2017
[35] W-H Lin Y-R Wang S-J Horng T-W Kao and Y Pan ldquoAblind watermarking method using maximum wavelet coeffi-cient quantizationrdquo Expert Systems with Applications vol 36no 9 pp 11509ndash11516 2009
[36] X Li J Liu J Sun X Yang andW Liu ldquoStep-projection-basedspread transform dither modulationrdquo IET Information Securityvol 5 no 3 pp 170ndash180 2011
[37] M Hamghalam S Mirzakuchaki andM A Akhaee ldquoGeomet-ric modelling of the wavelet coefficients for image watermark-ing using optimum detectorrdquo IET Image Processing vol 8 no 3pp 162ndash172 2014
[38] V S Verma R K Jha and A Ojha ldquoDigital watermarkextraction using support vector machine with principal com-ponent analysis based feature reductionrdquo Journal of VisualCommunication and Image Representation vol 31 pp 75ndash852015
[39] C Li Z Zhang Y Wang B Ma and D Huang ldquoDithermodulation of significant amplitude difference for waveletbased robust watermarkingrdquo Neurocomputing vol 166 pp404ndash415 2015
different keys and extract a watermark from a nonwater-marked image respectively In the two cases the maximumcorrelation obtained has not exceeded the value 059 whichpoints out a negative presence
5 Conclusion
The mean value of DWTrsquos vertical and horizontal subbandscoefficients is highly robust and imperceptible In fact these
two advantages were exploited in the design of a robustand blind watermarking approach The embedding methodconsisted in modulating the BMV in accordance with thequantization technique The use of a single transformationdomain alongside with a simple embedding technique makesfrom the proposed algorithmcomputationally undemandingAt the extraction phase the watermark bits were identifiedby the BMV position with respect to the zero value Unlikesome existing schemes the extraction rules did not require to
Security and Communication Networks 15
be adjusted to the applied attacks to fulfill a high robustnessIn fact the inserted watermark was capable to withstand withno error after JPEG compression up to a quality factor of 30Gaussian filter and other attacks
Data Availability
The data used to support the findings of this study areincluded within the article
Conflicts of Interest
The authors declare that they have no conflicts of interest
References
[1] S Liansheng Z Bei W Zhanmin and T Ailing ldquoAn opticalcolor image watermarking scheme by using compressive sens-ingwith human visual characteristics in gyrator domainrdquoOpticsand Lasers in Engineering vol 92 pp 85ndash93 2017
[2] D Shukla and M Sharma ldquoA Novel Scene-Based Video Water-marking Scheme for Copyright Protectionrdquo Journal of IntelligentSystems vol 27 no 1 pp 47ndash66 2018
[3] A Tiwari and M Sharma ldquoAn efficient vector quantizationbased watermarking method for image integrity authentica-tionrdquo Advances in Intelligent Systems and Computing vol 519pp 215ndash225 2018
[4] S Bravo-Solorio F Calderon C-T Li and A K Nandi ldquoFastfragile watermark embedding and iterative mechanism withhigh self-restoration performancerdquo Digital Signal Processingvol 73 pp 83ndash92 2018
[5] A Ghardallou A Kricha A Sakly and A Mtibaa ldquoAdaptiveblock sized reversible watermarking scheme based on integertransformrdquo in Proceedings of the 17th International Conferenceon Sciences and Techniques of Automatic Control and ComputerEngineering STA 2016 pp 347ndash351 Tunisia December 2016
[6] R G van Schyndel A Z Tirkel and C F Osborne ldquoA digitalwatermarkrdquo in Proceedings of the IEEE International ConferenceImage Processing (ICIP rsquo94) vol 2 pp 86ndash90 Austin Tex USANovember 1994
[7] W-H Lin S-J Horng T-W Kao P Fan C-L Lee and Y PanldquoAn efficient watermarking method based on significant differ-ence of wavelet coefficient quantizationrdquo IEEE Transactions onMultimedia vol 10 no 5 pp 746ndash757 2008
[8] S Bhattacharyya and G Sanyal ldquoA Robust Image Steganogra-phy using DWT Difference Modulation (DWTDM)rdquo Interna-tional Journal of Computer Network and Information Securityvol 4 no 7 pp 27ndash40 2012
[9] H S Malvar and D A Florencio ldquoImproved spread spectruma new modulation technique for robust watermarkingrdquo IEEETransactions on Signal Processing vol 51 no 4 pp 898ndash9052003
[10] B Chen and G W Wornell ldquoProvably robust digital water-markingrdquo in Proceedings of the 1999 Multimedia Systems andApplications II pp 43ndash54 September 1999
[11] Z Kricha A Kricha and A Sakly ldquoAccommodative extractorfor QIM-based watermarking schemesrdquo IET Image Processing2018
[12] I J Cox J Kilian T Leighton and T Shamoon ldquoSecure imper-ceptable yet perceptually salient spread spectrum watermark
formultimediardquo in Proceedings of the 1996 Southcon Conferencepp 192ndash197 June 1996
[13] X Li J Liu J Sun Z Sun H Hu and L Zhang ldquoImprovedspread transform dither modulation-based watermark algo-rithm based on step projectionrdquo in Proceedings of the 2010 IEEEInternational Symposium on Broadband Multimedia Systemsand Broadcasting (BMSB) pp 1ndash5 Shanghai China March2010
[14] H-T Hu and L-Y Hsu ldquoExploring DWT-SVD-DCT featureparameters for robust multiple watermarking against JPEG andJPEG2000 compressionrdquo Computers and Electrical Engineeringvol 41 pp 52ndash63 2015
[15] J Cannons and P Moulin ldquoDesign and statistical analysis of ahash-aided image watermarking systemrdquo IEEE Transactions onImage Processing vol 13 no 10 pp 1393ndash1408 2004
[16] A Valizadeh and Z J Wang ldquoAn improved multiplicativespread spectrum embedding scheme for data hidingrdquo IEEETransactions on Information Forensics and Security vol 7 no4 pp 1127ndash1143 2012
[17] M Kutter D Frederic and F B JordanDigital signature of colorimages using amplitude modulation 1997
[18] Q Su and B Chen ldquoRobust color image watermarking tech-nique in the spatial domainrdquo Soft Computing vol 22 no 1 pp91ndash106 2018
[19] Z Yanling Z Xiaoshi L Na et al ldquoA Digital ImageWatermarkAlgorithm Based onDCCoefficientsQuantizationrdquo in Proceed-ings of the 2006 6th World Congress on Intelligent Control andAutomation pp 9734ndash9738 Dalian China 2006
[20] RMZhaoH LianHWPang andBNHu ldquoAWatermarkingAlgorithm by Modifying AC Coefficies in DCT Domainrdquo inProceedings of the 2008 International Symposium on InformationScience and Engineering (ISISE) pp 159ndash162 Shanghai ChinaDecember 2008
[21] B L Gunjal and R RManthalkar ldquoDiscreteWavelet Transformbased strongly robust watermarking scheme for informationhiding in digital imagesrdquo in Proceedings of the 3rd InternationalConference on Emerging Trends in Engineering and TechnologyICETET 2010 pp 124ndash129 India November 2010
[22] H Shi and F Lv ldquoA Blind Digital Watermark Techniquefor Color Image Based on Integer Wavelet Transformrdquo inProceedings of the 2010 International Conference on BiomedicalEngineering and Computer Science (ICBECS) pp 1ndash4 WuhanChina April 2010
[23] S Ranjbar F Zargari and M Ghanbari ldquoA highly robust two-stage Contourlet-based digital image watermarking methodrdquoSignal Processing Image Communication vol 28 no 10 pp1526ndash1536 2013
[24] Q Su and B Chen ldquoA novel blind color image watermarkingusing upper Hessenbergmatrixrdquo AEU - International Journal ofElectronics and Communications vol 78 pp 64ndash71 2017
[25] F Ourique V Licks R Jordan and F Perez-Gonzalez ldquoAngleQIM a novel watermark embedding scheme robust againstamplitude scaling distortionsrdquo in Proceedings of the IEEE Inter-national Conference on Acoustics Speech and Signal Processing(ICASSP rsquo05) pp 797ndash800 March 2005
[26] H R Fazlali S Samavi N Karimi and S Shirani ldquoAdaptiveblind image watermarking using edge pixel concentrationrdquoMultimedia Tools and Applications vol 76 no 2 pp 3105ndash31202017
[27] X-Y Wang Y-N Liu H Xu A-L Wang and H-Y YangldquoBlind optimum detector for robust image watermarking in
16 Security and Communication Networks
nonsubsampled shearlet Domainrdquo Information Sciences vol372 pp 634ndash654 2016
[28] X Liu C Lin and S Yuan ldquoBlind Dual Watermarking forColor Imagesrsquo Authentication and Copyright Protectionrdquo IEEETransactions on Circuits and Systems for Video Technology vol28 no 5 pp 1047ndash1055 2018
[29] Poonam and S M Arora ldquoA DWT-SVD based Robust DigitalWatermarking for Digital Imagesrdquo Procedia Computer Sciencevol 132 pp 1441ndash1448 2018
[30] S Fazli and M Moeini ldquoA robust image watermarking methodbased on DWT DCT and SVD using a new technique forcorrection of main geometric attacksrdquo Optik - InternationalJournal for Light and ElectronOptics vol 127 no 2 pp 964ndash9722016
[31] K Kuppusamy and K Thamodaran ldquoOptimized Image Water-marking Scheme Based On PSOrdquo Procedia Engineering vol 38pp 493ndash503 2012
[32] V S Rao R S Shekhawat and V K Srivastava ldquoA dwt-dct-svd based digital image watermarking scheme using particleswarm optimizationrdquo in Proceedings of the 2012 IEEE StudentsrsquoConference on Electrical Electronics and Computer Science pp1ndash4 2012
[33] T T Takore P R Kumar and G L Devi ldquoA modified blindimage watermarking scheme based on DWT DCT and SVDdomain using GA to optimize robustnessrdquo in Proceedings ofthe 2016 International Conference on Electrical Electronics andOptimization Techniques ICEEOT 2016 pp 2725ndash2729 IndiaMarch 2016
[34] H Hu S Lin and L Hsu ldquoEffective blind speech watermarkingvia adaptive mean modulation and package synchronization inDWT domainrdquo EURASIP Journal on Audio Speech and MusicProcessing vol 2017 no 1 2017
[35] W-H Lin Y-R Wang S-J Horng T-W Kao and Y Pan ldquoAblind watermarking method using maximum wavelet coeffi-cient quantizationrdquo Expert Systems with Applications vol 36no 9 pp 11509ndash11516 2009
[36] X Li J Liu J Sun X Yang andW Liu ldquoStep-projection-basedspread transform dither modulationrdquo IET Information Securityvol 5 no 3 pp 170ndash180 2011
[37] M Hamghalam S Mirzakuchaki andM A Akhaee ldquoGeomet-ric modelling of the wavelet coefficients for image watermark-ing using optimum detectorrdquo IET Image Processing vol 8 no 3pp 162ndash172 2014
[38] V S Verma R K Jha and A Ojha ldquoDigital watermarkextraction using support vector machine with principal com-ponent analysis based feature reductionrdquo Journal of VisualCommunication and Image Representation vol 31 pp 75ndash852015
[39] C Li Z Zhang Y Wang B Ma and D Huang ldquoDithermodulation of significant amplitude difference for waveletbased robust watermarkingrdquo Neurocomputing vol 166 pp404ndash415 2015
different keys and extract a watermark from a nonwater-marked image respectively In the two cases the maximumcorrelation obtained has not exceeded the value 059 whichpoints out a negative presence
5 Conclusion
The mean value of DWTrsquos vertical and horizontal subbandscoefficients is highly robust and imperceptible In fact these
two advantages were exploited in the design of a robustand blind watermarking approach The embedding methodconsisted in modulating the BMV in accordance with thequantization technique The use of a single transformationdomain alongside with a simple embedding technique makesfrom the proposed algorithmcomputationally undemandingAt the extraction phase the watermark bits were identifiedby the BMV position with respect to the zero value Unlikesome existing schemes the extraction rules did not require to
Security and Communication Networks 15
be adjusted to the applied attacks to fulfill a high robustnessIn fact the inserted watermark was capable to withstand withno error after JPEG compression up to a quality factor of 30Gaussian filter and other attacks
Data Availability
The data used to support the findings of this study areincluded within the article
Conflicts of Interest
The authors declare that they have no conflicts of interest
References
[1] S Liansheng Z Bei W Zhanmin and T Ailing ldquoAn opticalcolor image watermarking scheme by using compressive sens-ingwith human visual characteristics in gyrator domainrdquoOpticsand Lasers in Engineering vol 92 pp 85ndash93 2017
[2] D Shukla and M Sharma ldquoA Novel Scene-Based Video Water-marking Scheme for Copyright Protectionrdquo Journal of IntelligentSystems vol 27 no 1 pp 47ndash66 2018
[3] A Tiwari and M Sharma ldquoAn efficient vector quantizationbased watermarking method for image integrity authentica-tionrdquo Advances in Intelligent Systems and Computing vol 519pp 215ndash225 2018
[4] S Bravo-Solorio F Calderon C-T Li and A K Nandi ldquoFastfragile watermark embedding and iterative mechanism withhigh self-restoration performancerdquo Digital Signal Processingvol 73 pp 83ndash92 2018
[5] A Ghardallou A Kricha A Sakly and A Mtibaa ldquoAdaptiveblock sized reversible watermarking scheme based on integertransformrdquo in Proceedings of the 17th International Conferenceon Sciences and Techniques of Automatic Control and ComputerEngineering STA 2016 pp 347ndash351 Tunisia December 2016
[6] R G van Schyndel A Z Tirkel and C F Osborne ldquoA digitalwatermarkrdquo in Proceedings of the IEEE International ConferenceImage Processing (ICIP rsquo94) vol 2 pp 86ndash90 Austin Tex USANovember 1994
[7] W-H Lin S-J Horng T-W Kao P Fan C-L Lee and Y PanldquoAn efficient watermarking method based on significant differ-ence of wavelet coefficient quantizationrdquo IEEE Transactions onMultimedia vol 10 no 5 pp 746ndash757 2008
[8] S Bhattacharyya and G Sanyal ldquoA Robust Image Steganogra-phy using DWT Difference Modulation (DWTDM)rdquo Interna-tional Journal of Computer Network and Information Securityvol 4 no 7 pp 27ndash40 2012
[9] H S Malvar and D A Florencio ldquoImproved spread spectruma new modulation technique for robust watermarkingrdquo IEEETransactions on Signal Processing vol 51 no 4 pp 898ndash9052003
[10] B Chen and G W Wornell ldquoProvably robust digital water-markingrdquo in Proceedings of the 1999 Multimedia Systems andApplications II pp 43ndash54 September 1999
[11] Z Kricha A Kricha and A Sakly ldquoAccommodative extractorfor QIM-based watermarking schemesrdquo IET Image Processing2018
[12] I J Cox J Kilian T Leighton and T Shamoon ldquoSecure imper-ceptable yet perceptually salient spread spectrum watermark
formultimediardquo in Proceedings of the 1996 Southcon Conferencepp 192ndash197 June 1996
[13] X Li J Liu J Sun Z Sun H Hu and L Zhang ldquoImprovedspread transform dither modulation-based watermark algo-rithm based on step projectionrdquo in Proceedings of the 2010 IEEEInternational Symposium on Broadband Multimedia Systemsand Broadcasting (BMSB) pp 1ndash5 Shanghai China March2010
[14] H-T Hu and L-Y Hsu ldquoExploring DWT-SVD-DCT featureparameters for robust multiple watermarking against JPEG andJPEG2000 compressionrdquo Computers and Electrical Engineeringvol 41 pp 52ndash63 2015
[15] J Cannons and P Moulin ldquoDesign and statistical analysis of ahash-aided image watermarking systemrdquo IEEE Transactions onImage Processing vol 13 no 10 pp 1393ndash1408 2004
[16] A Valizadeh and Z J Wang ldquoAn improved multiplicativespread spectrum embedding scheme for data hidingrdquo IEEETransactions on Information Forensics and Security vol 7 no4 pp 1127ndash1143 2012
[17] M Kutter D Frederic and F B JordanDigital signature of colorimages using amplitude modulation 1997
[18] Q Su and B Chen ldquoRobust color image watermarking tech-nique in the spatial domainrdquo Soft Computing vol 22 no 1 pp91ndash106 2018
[19] Z Yanling Z Xiaoshi L Na et al ldquoA Digital ImageWatermarkAlgorithm Based onDCCoefficientsQuantizationrdquo in Proceed-ings of the 2006 6th World Congress on Intelligent Control andAutomation pp 9734ndash9738 Dalian China 2006
[20] RMZhaoH LianHWPang andBNHu ldquoAWatermarkingAlgorithm by Modifying AC Coefficies in DCT Domainrdquo inProceedings of the 2008 International Symposium on InformationScience and Engineering (ISISE) pp 159ndash162 Shanghai ChinaDecember 2008
[21] B L Gunjal and R RManthalkar ldquoDiscreteWavelet Transformbased strongly robust watermarking scheme for informationhiding in digital imagesrdquo in Proceedings of the 3rd InternationalConference on Emerging Trends in Engineering and TechnologyICETET 2010 pp 124ndash129 India November 2010
[22] H Shi and F Lv ldquoA Blind Digital Watermark Techniquefor Color Image Based on Integer Wavelet Transformrdquo inProceedings of the 2010 International Conference on BiomedicalEngineering and Computer Science (ICBECS) pp 1ndash4 WuhanChina April 2010
[23] S Ranjbar F Zargari and M Ghanbari ldquoA highly robust two-stage Contourlet-based digital image watermarking methodrdquoSignal Processing Image Communication vol 28 no 10 pp1526ndash1536 2013
[24] Q Su and B Chen ldquoA novel blind color image watermarkingusing upper Hessenbergmatrixrdquo AEU - International Journal ofElectronics and Communications vol 78 pp 64ndash71 2017
[25] F Ourique V Licks R Jordan and F Perez-Gonzalez ldquoAngleQIM a novel watermark embedding scheme robust againstamplitude scaling distortionsrdquo in Proceedings of the IEEE Inter-national Conference on Acoustics Speech and Signal Processing(ICASSP rsquo05) pp 797ndash800 March 2005
[26] H R Fazlali S Samavi N Karimi and S Shirani ldquoAdaptiveblind image watermarking using edge pixel concentrationrdquoMultimedia Tools and Applications vol 76 no 2 pp 3105ndash31202017
[27] X-Y Wang Y-N Liu H Xu A-L Wang and H-Y YangldquoBlind optimum detector for robust image watermarking in
16 Security and Communication Networks
nonsubsampled shearlet Domainrdquo Information Sciences vol372 pp 634ndash654 2016
[28] X Liu C Lin and S Yuan ldquoBlind Dual Watermarking forColor Imagesrsquo Authentication and Copyright Protectionrdquo IEEETransactions on Circuits and Systems for Video Technology vol28 no 5 pp 1047ndash1055 2018
[29] Poonam and S M Arora ldquoA DWT-SVD based Robust DigitalWatermarking for Digital Imagesrdquo Procedia Computer Sciencevol 132 pp 1441ndash1448 2018
[30] S Fazli and M Moeini ldquoA robust image watermarking methodbased on DWT DCT and SVD using a new technique forcorrection of main geometric attacksrdquo Optik - InternationalJournal for Light and ElectronOptics vol 127 no 2 pp 964ndash9722016
[31] K Kuppusamy and K Thamodaran ldquoOptimized Image Water-marking Scheme Based On PSOrdquo Procedia Engineering vol 38pp 493ndash503 2012
[32] V S Rao R S Shekhawat and V K Srivastava ldquoA dwt-dct-svd based digital image watermarking scheme using particleswarm optimizationrdquo in Proceedings of the 2012 IEEE StudentsrsquoConference on Electrical Electronics and Computer Science pp1ndash4 2012
[33] T T Takore P R Kumar and G L Devi ldquoA modified blindimage watermarking scheme based on DWT DCT and SVDdomain using GA to optimize robustnessrdquo in Proceedings ofthe 2016 International Conference on Electrical Electronics andOptimization Techniques ICEEOT 2016 pp 2725ndash2729 IndiaMarch 2016
[34] H Hu S Lin and L Hsu ldquoEffective blind speech watermarkingvia adaptive mean modulation and package synchronization inDWT domainrdquo EURASIP Journal on Audio Speech and MusicProcessing vol 2017 no 1 2017
[35] W-H Lin Y-R Wang S-J Horng T-W Kao and Y Pan ldquoAblind watermarking method using maximum wavelet coeffi-cient quantizationrdquo Expert Systems with Applications vol 36no 9 pp 11509ndash11516 2009
[36] X Li J Liu J Sun X Yang andW Liu ldquoStep-projection-basedspread transform dither modulationrdquo IET Information Securityvol 5 no 3 pp 170ndash180 2011
[37] M Hamghalam S Mirzakuchaki andM A Akhaee ldquoGeomet-ric modelling of the wavelet coefficients for image watermark-ing using optimum detectorrdquo IET Image Processing vol 8 no 3pp 162ndash172 2014
[38] V S Verma R K Jha and A Ojha ldquoDigital watermarkextraction using support vector machine with principal com-ponent analysis based feature reductionrdquo Journal of VisualCommunication and Image Representation vol 31 pp 75ndash852015
[39] C Li Z Zhang Y Wang B Ma and D Huang ldquoDithermodulation of significant amplitude difference for waveletbased robust watermarkingrdquo Neurocomputing vol 166 pp404ndash415 2015
different keys and extract a watermark from a nonwater-marked image respectively In the two cases the maximumcorrelation obtained has not exceeded the value 059 whichpoints out a negative presence
5 Conclusion
The mean value of DWTrsquos vertical and horizontal subbandscoefficients is highly robust and imperceptible In fact these
two advantages were exploited in the design of a robustand blind watermarking approach The embedding methodconsisted in modulating the BMV in accordance with thequantization technique The use of a single transformationdomain alongside with a simple embedding technique makesfrom the proposed algorithmcomputationally undemandingAt the extraction phase the watermark bits were identifiedby the BMV position with respect to the zero value Unlikesome existing schemes the extraction rules did not require to
Security and Communication Networks 15
be adjusted to the applied attacks to fulfill a high robustnessIn fact the inserted watermark was capable to withstand withno error after JPEG compression up to a quality factor of 30Gaussian filter and other attacks
Data Availability
The data used to support the findings of this study areincluded within the article
Conflicts of Interest
The authors declare that they have no conflicts of interest
References
[1] S Liansheng Z Bei W Zhanmin and T Ailing ldquoAn opticalcolor image watermarking scheme by using compressive sens-ingwith human visual characteristics in gyrator domainrdquoOpticsand Lasers in Engineering vol 92 pp 85ndash93 2017
[2] D Shukla and M Sharma ldquoA Novel Scene-Based Video Water-marking Scheme for Copyright Protectionrdquo Journal of IntelligentSystems vol 27 no 1 pp 47ndash66 2018
[3] A Tiwari and M Sharma ldquoAn efficient vector quantizationbased watermarking method for image integrity authentica-tionrdquo Advances in Intelligent Systems and Computing vol 519pp 215ndash225 2018
[4] S Bravo-Solorio F Calderon C-T Li and A K Nandi ldquoFastfragile watermark embedding and iterative mechanism withhigh self-restoration performancerdquo Digital Signal Processingvol 73 pp 83ndash92 2018
[5] A Ghardallou A Kricha A Sakly and A Mtibaa ldquoAdaptiveblock sized reversible watermarking scheme based on integertransformrdquo in Proceedings of the 17th International Conferenceon Sciences and Techniques of Automatic Control and ComputerEngineering STA 2016 pp 347ndash351 Tunisia December 2016
[6] R G van Schyndel A Z Tirkel and C F Osborne ldquoA digitalwatermarkrdquo in Proceedings of the IEEE International ConferenceImage Processing (ICIP rsquo94) vol 2 pp 86ndash90 Austin Tex USANovember 1994
[7] W-H Lin S-J Horng T-W Kao P Fan C-L Lee and Y PanldquoAn efficient watermarking method based on significant differ-ence of wavelet coefficient quantizationrdquo IEEE Transactions onMultimedia vol 10 no 5 pp 746ndash757 2008
[8] S Bhattacharyya and G Sanyal ldquoA Robust Image Steganogra-phy using DWT Difference Modulation (DWTDM)rdquo Interna-tional Journal of Computer Network and Information Securityvol 4 no 7 pp 27ndash40 2012
[9] H S Malvar and D A Florencio ldquoImproved spread spectruma new modulation technique for robust watermarkingrdquo IEEETransactions on Signal Processing vol 51 no 4 pp 898ndash9052003
[10] B Chen and G W Wornell ldquoProvably robust digital water-markingrdquo in Proceedings of the 1999 Multimedia Systems andApplications II pp 43ndash54 September 1999
[11] Z Kricha A Kricha and A Sakly ldquoAccommodative extractorfor QIM-based watermarking schemesrdquo IET Image Processing2018
[12] I J Cox J Kilian T Leighton and T Shamoon ldquoSecure imper-ceptable yet perceptually salient spread spectrum watermark
formultimediardquo in Proceedings of the 1996 Southcon Conferencepp 192ndash197 June 1996
[13] X Li J Liu J Sun Z Sun H Hu and L Zhang ldquoImprovedspread transform dither modulation-based watermark algo-rithm based on step projectionrdquo in Proceedings of the 2010 IEEEInternational Symposium on Broadband Multimedia Systemsand Broadcasting (BMSB) pp 1ndash5 Shanghai China March2010
[14] H-T Hu and L-Y Hsu ldquoExploring DWT-SVD-DCT featureparameters for robust multiple watermarking against JPEG andJPEG2000 compressionrdquo Computers and Electrical Engineeringvol 41 pp 52ndash63 2015
[15] J Cannons and P Moulin ldquoDesign and statistical analysis of ahash-aided image watermarking systemrdquo IEEE Transactions onImage Processing vol 13 no 10 pp 1393ndash1408 2004
[16] A Valizadeh and Z J Wang ldquoAn improved multiplicativespread spectrum embedding scheme for data hidingrdquo IEEETransactions on Information Forensics and Security vol 7 no4 pp 1127ndash1143 2012
[17] M Kutter D Frederic and F B JordanDigital signature of colorimages using amplitude modulation 1997
[18] Q Su and B Chen ldquoRobust color image watermarking tech-nique in the spatial domainrdquo Soft Computing vol 22 no 1 pp91ndash106 2018
[19] Z Yanling Z Xiaoshi L Na et al ldquoA Digital ImageWatermarkAlgorithm Based onDCCoefficientsQuantizationrdquo in Proceed-ings of the 2006 6th World Congress on Intelligent Control andAutomation pp 9734ndash9738 Dalian China 2006
[20] RMZhaoH LianHWPang andBNHu ldquoAWatermarkingAlgorithm by Modifying AC Coefficies in DCT Domainrdquo inProceedings of the 2008 International Symposium on InformationScience and Engineering (ISISE) pp 159ndash162 Shanghai ChinaDecember 2008
[21] B L Gunjal and R RManthalkar ldquoDiscreteWavelet Transformbased strongly robust watermarking scheme for informationhiding in digital imagesrdquo in Proceedings of the 3rd InternationalConference on Emerging Trends in Engineering and TechnologyICETET 2010 pp 124ndash129 India November 2010
[22] H Shi and F Lv ldquoA Blind Digital Watermark Techniquefor Color Image Based on Integer Wavelet Transformrdquo inProceedings of the 2010 International Conference on BiomedicalEngineering and Computer Science (ICBECS) pp 1ndash4 WuhanChina April 2010
[23] S Ranjbar F Zargari and M Ghanbari ldquoA highly robust two-stage Contourlet-based digital image watermarking methodrdquoSignal Processing Image Communication vol 28 no 10 pp1526ndash1536 2013
[24] Q Su and B Chen ldquoA novel blind color image watermarkingusing upper Hessenbergmatrixrdquo AEU - International Journal ofElectronics and Communications vol 78 pp 64ndash71 2017
[25] F Ourique V Licks R Jordan and F Perez-Gonzalez ldquoAngleQIM a novel watermark embedding scheme robust againstamplitude scaling distortionsrdquo in Proceedings of the IEEE Inter-national Conference on Acoustics Speech and Signal Processing(ICASSP rsquo05) pp 797ndash800 March 2005
[26] H R Fazlali S Samavi N Karimi and S Shirani ldquoAdaptiveblind image watermarking using edge pixel concentrationrdquoMultimedia Tools and Applications vol 76 no 2 pp 3105ndash31202017
[27] X-Y Wang Y-N Liu H Xu A-L Wang and H-Y YangldquoBlind optimum detector for robust image watermarking in
16 Security and Communication Networks
nonsubsampled shearlet Domainrdquo Information Sciences vol372 pp 634ndash654 2016
[28] X Liu C Lin and S Yuan ldquoBlind Dual Watermarking forColor Imagesrsquo Authentication and Copyright Protectionrdquo IEEETransactions on Circuits and Systems for Video Technology vol28 no 5 pp 1047ndash1055 2018
[29] Poonam and S M Arora ldquoA DWT-SVD based Robust DigitalWatermarking for Digital Imagesrdquo Procedia Computer Sciencevol 132 pp 1441ndash1448 2018
[30] S Fazli and M Moeini ldquoA robust image watermarking methodbased on DWT DCT and SVD using a new technique forcorrection of main geometric attacksrdquo Optik - InternationalJournal for Light and ElectronOptics vol 127 no 2 pp 964ndash9722016
[31] K Kuppusamy and K Thamodaran ldquoOptimized Image Water-marking Scheme Based On PSOrdquo Procedia Engineering vol 38pp 493ndash503 2012
[32] V S Rao R S Shekhawat and V K Srivastava ldquoA dwt-dct-svd based digital image watermarking scheme using particleswarm optimizationrdquo in Proceedings of the 2012 IEEE StudentsrsquoConference on Electrical Electronics and Computer Science pp1ndash4 2012
[33] T T Takore P R Kumar and G L Devi ldquoA modified blindimage watermarking scheme based on DWT DCT and SVDdomain using GA to optimize robustnessrdquo in Proceedings ofthe 2016 International Conference on Electrical Electronics andOptimization Techniques ICEEOT 2016 pp 2725ndash2729 IndiaMarch 2016
[34] H Hu S Lin and L Hsu ldquoEffective blind speech watermarkingvia adaptive mean modulation and package synchronization inDWT domainrdquo EURASIP Journal on Audio Speech and MusicProcessing vol 2017 no 1 2017
[35] W-H Lin Y-R Wang S-J Horng T-W Kao and Y Pan ldquoAblind watermarking method using maximum wavelet coeffi-cient quantizationrdquo Expert Systems with Applications vol 36no 9 pp 11509ndash11516 2009
[36] X Li J Liu J Sun X Yang andW Liu ldquoStep-projection-basedspread transform dither modulationrdquo IET Information Securityvol 5 no 3 pp 170ndash180 2011
[37] M Hamghalam S Mirzakuchaki andM A Akhaee ldquoGeomet-ric modelling of the wavelet coefficients for image watermark-ing using optimum detectorrdquo IET Image Processing vol 8 no 3pp 162ndash172 2014
[38] V S Verma R K Jha and A Ojha ldquoDigital watermarkextraction using support vector machine with principal com-ponent analysis based feature reductionrdquo Journal of VisualCommunication and Image Representation vol 31 pp 75ndash852015
[39] C Li Z Zhang Y Wang B Ma and D Huang ldquoDithermodulation of significant amplitude difference for waveletbased robust watermarkingrdquo Neurocomputing vol 166 pp404ndash415 2015
different keys and extract a watermark from a nonwater-marked image respectively In the two cases the maximumcorrelation obtained has not exceeded the value 059 whichpoints out a negative presence
5 Conclusion
The mean value of DWTrsquos vertical and horizontal subbandscoefficients is highly robust and imperceptible In fact these
two advantages were exploited in the design of a robustand blind watermarking approach The embedding methodconsisted in modulating the BMV in accordance with thequantization technique The use of a single transformationdomain alongside with a simple embedding technique makesfrom the proposed algorithmcomputationally undemandingAt the extraction phase the watermark bits were identifiedby the BMV position with respect to the zero value Unlikesome existing schemes the extraction rules did not require to
Security and Communication Networks 15
be adjusted to the applied attacks to fulfill a high robustnessIn fact the inserted watermark was capable to withstand withno error after JPEG compression up to a quality factor of 30Gaussian filter and other attacks
Data Availability
The data used to support the findings of this study areincluded within the article
Conflicts of Interest
The authors declare that they have no conflicts of interest
References
[1] S Liansheng Z Bei W Zhanmin and T Ailing ldquoAn opticalcolor image watermarking scheme by using compressive sens-ingwith human visual characteristics in gyrator domainrdquoOpticsand Lasers in Engineering vol 92 pp 85ndash93 2017
[2] D Shukla and M Sharma ldquoA Novel Scene-Based Video Water-marking Scheme for Copyright Protectionrdquo Journal of IntelligentSystems vol 27 no 1 pp 47ndash66 2018
[3] A Tiwari and M Sharma ldquoAn efficient vector quantizationbased watermarking method for image integrity authentica-tionrdquo Advances in Intelligent Systems and Computing vol 519pp 215ndash225 2018
[4] S Bravo-Solorio F Calderon C-T Li and A K Nandi ldquoFastfragile watermark embedding and iterative mechanism withhigh self-restoration performancerdquo Digital Signal Processingvol 73 pp 83ndash92 2018
[5] A Ghardallou A Kricha A Sakly and A Mtibaa ldquoAdaptiveblock sized reversible watermarking scheme based on integertransformrdquo in Proceedings of the 17th International Conferenceon Sciences and Techniques of Automatic Control and ComputerEngineering STA 2016 pp 347ndash351 Tunisia December 2016
[6] R G van Schyndel A Z Tirkel and C F Osborne ldquoA digitalwatermarkrdquo in Proceedings of the IEEE International ConferenceImage Processing (ICIP rsquo94) vol 2 pp 86ndash90 Austin Tex USANovember 1994
[7] W-H Lin S-J Horng T-W Kao P Fan C-L Lee and Y PanldquoAn efficient watermarking method based on significant differ-ence of wavelet coefficient quantizationrdquo IEEE Transactions onMultimedia vol 10 no 5 pp 746ndash757 2008
[8] S Bhattacharyya and G Sanyal ldquoA Robust Image Steganogra-phy using DWT Difference Modulation (DWTDM)rdquo Interna-tional Journal of Computer Network and Information Securityvol 4 no 7 pp 27ndash40 2012
[9] H S Malvar and D A Florencio ldquoImproved spread spectruma new modulation technique for robust watermarkingrdquo IEEETransactions on Signal Processing vol 51 no 4 pp 898ndash9052003
[10] B Chen and G W Wornell ldquoProvably robust digital water-markingrdquo in Proceedings of the 1999 Multimedia Systems andApplications II pp 43ndash54 September 1999
[11] Z Kricha A Kricha and A Sakly ldquoAccommodative extractorfor QIM-based watermarking schemesrdquo IET Image Processing2018
[12] I J Cox J Kilian T Leighton and T Shamoon ldquoSecure imper-ceptable yet perceptually salient spread spectrum watermark
formultimediardquo in Proceedings of the 1996 Southcon Conferencepp 192ndash197 June 1996
[13] X Li J Liu J Sun Z Sun H Hu and L Zhang ldquoImprovedspread transform dither modulation-based watermark algo-rithm based on step projectionrdquo in Proceedings of the 2010 IEEEInternational Symposium on Broadband Multimedia Systemsand Broadcasting (BMSB) pp 1ndash5 Shanghai China March2010
[14] H-T Hu and L-Y Hsu ldquoExploring DWT-SVD-DCT featureparameters for robust multiple watermarking against JPEG andJPEG2000 compressionrdquo Computers and Electrical Engineeringvol 41 pp 52ndash63 2015
[15] J Cannons and P Moulin ldquoDesign and statistical analysis of ahash-aided image watermarking systemrdquo IEEE Transactions onImage Processing vol 13 no 10 pp 1393ndash1408 2004
[16] A Valizadeh and Z J Wang ldquoAn improved multiplicativespread spectrum embedding scheme for data hidingrdquo IEEETransactions on Information Forensics and Security vol 7 no4 pp 1127ndash1143 2012
[17] M Kutter D Frederic and F B JordanDigital signature of colorimages using amplitude modulation 1997
[18] Q Su and B Chen ldquoRobust color image watermarking tech-nique in the spatial domainrdquo Soft Computing vol 22 no 1 pp91ndash106 2018
[19] Z Yanling Z Xiaoshi L Na et al ldquoA Digital ImageWatermarkAlgorithm Based onDCCoefficientsQuantizationrdquo in Proceed-ings of the 2006 6th World Congress on Intelligent Control andAutomation pp 9734ndash9738 Dalian China 2006
[20] RMZhaoH LianHWPang andBNHu ldquoAWatermarkingAlgorithm by Modifying AC Coefficies in DCT Domainrdquo inProceedings of the 2008 International Symposium on InformationScience and Engineering (ISISE) pp 159ndash162 Shanghai ChinaDecember 2008
[21] B L Gunjal and R RManthalkar ldquoDiscreteWavelet Transformbased strongly robust watermarking scheme for informationhiding in digital imagesrdquo in Proceedings of the 3rd InternationalConference on Emerging Trends in Engineering and TechnologyICETET 2010 pp 124ndash129 India November 2010
[22] H Shi and F Lv ldquoA Blind Digital Watermark Techniquefor Color Image Based on Integer Wavelet Transformrdquo inProceedings of the 2010 International Conference on BiomedicalEngineering and Computer Science (ICBECS) pp 1ndash4 WuhanChina April 2010
[23] S Ranjbar F Zargari and M Ghanbari ldquoA highly robust two-stage Contourlet-based digital image watermarking methodrdquoSignal Processing Image Communication vol 28 no 10 pp1526ndash1536 2013
[24] Q Su and B Chen ldquoA novel blind color image watermarkingusing upper Hessenbergmatrixrdquo AEU - International Journal ofElectronics and Communications vol 78 pp 64ndash71 2017
[25] F Ourique V Licks R Jordan and F Perez-Gonzalez ldquoAngleQIM a novel watermark embedding scheme robust againstamplitude scaling distortionsrdquo in Proceedings of the IEEE Inter-national Conference on Acoustics Speech and Signal Processing(ICASSP rsquo05) pp 797ndash800 March 2005
[26] H R Fazlali S Samavi N Karimi and S Shirani ldquoAdaptiveblind image watermarking using edge pixel concentrationrdquoMultimedia Tools and Applications vol 76 no 2 pp 3105ndash31202017
[27] X-Y Wang Y-N Liu H Xu A-L Wang and H-Y YangldquoBlind optimum detector for robust image watermarking in
16 Security and Communication Networks
nonsubsampled shearlet Domainrdquo Information Sciences vol372 pp 634ndash654 2016
[28] X Liu C Lin and S Yuan ldquoBlind Dual Watermarking forColor Imagesrsquo Authentication and Copyright Protectionrdquo IEEETransactions on Circuits and Systems for Video Technology vol28 no 5 pp 1047ndash1055 2018
[29] Poonam and S M Arora ldquoA DWT-SVD based Robust DigitalWatermarking for Digital Imagesrdquo Procedia Computer Sciencevol 132 pp 1441ndash1448 2018
[30] S Fazli and M Moeini ldquoA robust image watermarking methodbased on DWT DCT and SVD using a new technique forcorrection of main geometric attacksrdquo Optik - InternationalJournal for Light and ElectronOptics vol 127 no 2 pp 964ndash9722016
[31] K Kuppusamy and K Thamodaran ldquoOptimized Image Water-marking Scheme Based On PSOrdquo Procedia Engineering vol 38pp 493ndash503 2012
[32] V S Rao R S Shekhawat and V K Srivastava ldquoA dwt-dct-svd based digital image watermarking scheme using particleswarm optimizationrdquo in Proceedings of the 2012 IEEE StudentsrsquoConference on Electrical Electronics and Computer Science pp1ndash4 2012
[33] T T Takore P R Kumar and G L Devi ldquoA modified blindimage watermarking scheme based on DWT DCT and SVDdomain using GA to optimize robustnessrdquo in Proceedings ofthe 2016 International Conference on Electrical Electronics andOptimization Techniques ICEEOT 2016 pp 2725ndash2729 IndiaMarch 2016
[34] H Hu S Lin and L Hsu ldquoEffective blind speech watermarkingvia adaptive mean modulation and package synchronization inDWT domainrdquo EURASIP Journal on Audio Speech and MusicProcessing vol 2017 no 1 2017
[35] W-H Lin Y-R Wang S-J Horng T-W Kao and Y Pan ldquoAblind watermarking method using maximum wavelet coeffi-cient quantizationrdquo Expert Systems with Applications vol 36no 9 pp 11509ndash11516 2009
[36] X Li J Liu J Sun X Yang andW Liu ldquoStep-projection-basedspread transform dither modulationrdquo IET Information Securityvol 5 no 3 pp 170ndash180 2011
[37] M Hamghalam S Mirzakuchaki andM A Akhaee ldquoGeomet-ric modelling of the wavelet coefficients for image watermark-ing using optimum detectorrdquo IET Image Processing vol 8 no 3pp 162ndash172 2014
[38] V S Verma R K Jha and A Ojha ldquoDigital watermarkextraction using support vector machine with principal com-ponent analysis based feature reductionrdquo Journal of VisualCommunication and Image Representation vol 31 pp 75ndash852015
[39] C Li Z Zhang Y Wang B Ma and D Huang ldquoDithermodulation of significant amplitude difference for waveletbased robust watermarkingrdquo Neurocomputing vol 166 pp404ndash415 2015
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Navigation and Observation
International Journal of
Hindawi
wwwhindawicom Volume 2018
Advances in
Multimedia
Submit your manuscripts atwwwhindawicom
Security and Communication Networks 15
be adjusted to the applied attacks to fulfill a high robustnessIn fact the inserted watermark was capable to withstand withno error after JPEG compression up to a quality factor of 30Gaussian filter and other attacks
Data Availability
The data used to support the findings of this study areincluded within the article
Conflicts of Interest
The authors declare that they have no conflicts of interest
References
[1] S Liansheng Z Bei W Zhanmin and T Ailing ldquoAn opticalcolor image watermarking scheme by using compressive sens-ingwith human visual characteristics in gyrator domainrdquoOpticsand Lasers in Engineering vol 92 pp 85ndash93 2017
[2] D Shukla and M Sharma ldquoA Novel Scene-Based Video Water-marking Scheme for Copyright Protectionrdquo Journal of IntelligentSystems vol 27 no 1 pp 47ndash66 2018
[3] A Tiwari and M Sharma ldquoAn efficient vector quantizationbased watermarking method for image integrity authentica-tionrdquo Advances in Intelligent Systems and Computing vol 519pp 215ndash225 2018
[4] S Bravo-Solorio F Calderon C-T Li and A K Nandi ldquoFastfragile watermark embedding and iterative mechanism withhigh self-restoration performancerdquo Digital Signal Processingvol 73 pp 83ndash92 2018
[5] A Ghardallou A Kricha A Sakly and A Mtibaa ldquoAdaptiveblock sized reversible watermarking scheme based on integertransformrdquo in Proceedings of the 17th International Conferenceon Sciences and Techniques of Automatic Control and ComputerEngineering STA 2016 pp 347ndash351 Tunisia December 2016
[6] R G van Schyndel A Z Tirkel and C F Osborne ldquoA digitalwatermarkrdquo in Proceedings of the IEEE International ConferenceImage Processing (ICIP rsquo94) vol 2 pp 86ndash90 Austin Tex USANovember 1994
[7] W-H Lin S-J Horng T-W Kao P Fan C-L Lee and Y PanldquoAn efficient watermarking method based on significant differ-ence of wavelet coefficient quantizationrdquo IEEE Transactions onMultimedia vol 10 no 5 pp 746ndash757 2008
[8] S Bhattacharyya and G Sanyal ldquoA Robust Image Steganogra-phy using DWT Difference Modulation (DWTDM)rdquo Interna-tional Journal of Computer Network and Information Securityvol 4 no 7 pp 27ndash40 2012
[9] H S Malvar and D A Florencio ldquoImproved spread spectruma new modulation technique for robust watermarkingrdquo IEEETransactions on Signal Processing vol 51 no 4 pp 898ndash9052003
[10] B Chen and G W Wornell ldquoProvably robust digital water-markingrdquo in Proceedings of the 1999 Multimedia Systems andApplications II pp 43ndash54 September 1999
[11] Z Kricha A Kricha and A Sakly ldquoAccommodative extractorfor QIM-based watermarking schemesrdquo IET Image Processing2018
[12] I J Cox J Kilian T Leighton and T Shamoon ldquoSecure imper-ceptable yet perceptually salient spread spectrum watermark
formultimediardquo in Proceedings of the 1996 Southcon Conferencepp 192ndash197 June 1996
[13] X Li J Liu J Sun Z Sun H Hu and L Zhang ldquoImprovedspread transform dither modulation-based watermark algo-rithm based on step projectionrdquo in Proceedings of the 2010 IEEEInternational Symposium on Broadband Multimedia Systemsand Broadcasting (BMSB) pp 1ndash5 Shanghai China March2010
[14] H-T Hu and L-Y Hsu ldquoExploring DWT-SVD-DCT featureparameters for robust multiple watermarking against JPEG andJPEG2000 compressionrdquo Computers and Electrical Engineeringvol 41 pp 52ndash63 2015
[15] J Cannons and P Moulin ldquoDesign and statistical analysis of ahash-aided image watermarking systemrdquo IEEE Transactions onImage Processing vol 13 no 10 pp 1393ndash1408 2004
[16] A Valizadeh and Z J Wang ldquoAn improved multiplicativespread spectrum embedding scheme for data hidingrdquo IEEETransactions on Information Forensics and Security vol 7 no4 pp 1127ndash1143 2012
[17] M Kutter D Frederic and F B JordanDigital signature of colorimages using amplitude modulation 1997
[18] Q Su and B Chen ldquoRobust color image watermarking tech-nique in the spatial domainrdquo Soft Computing vol 22 no 1 pp91ndash106 2018
[19] Z Yanling Z Xiaoshi L Na et al ldquoA Digital ImageWatermarkAlgorithm Based onDCCoefficientsQuantizationrdquo in Proceed-ings of the 2006 6th World Congress on Intelligent Control andAutomation pp 9734ndash9738 Dalian China 2006
[20] RMZhaoH LianHWPang andBNHu ldquoAWatermarkingAlgorithm by Modifying AC Coefficies in DCT Domainrdquo inProceedings of the 2008 International Symposium on InformationScience and Engineering (ISISE) pp 159ndash162 Shanghai ChinaDecember 2008
[21] B L Gunjal and R RManthalkar ldquoDiscreteWavelet Transformbased strongly robust watermarking scheme for informationhiding in digital imagesrdquo in Proceedings of the 3rd InternationalConference on Emerging Trends in Engineering and TechnologyICETET 2010 pp 124ndash129 India November 2010
[22] H Shi and F Lv ldquoA Blind Digital Watermark Techniquefor Color Image Based on Integer Wavelet Transformrdquo inProceedings of the 2010 International Conference on BiomedicalEngineering and Computer Science (ICBECS) pp 1ndash4 WuhanChina April 2010
[23] S Ranjbar F Zargari and M Ghanbari ldquoA highly robust two-stage Contourlet-based digital image watermarking methodrdquoSignal Processing Image Communication vol 28 no 10 pp1526ndash1536 2013
[24] Q Su and B Chen ldquoA novel blind color image watermarkingusing upper Hessenbergmatrixrdquo AEU - International Journal ofElectronics and Communications vol 78 pp 64ndash71 2017
[25] F Ourique V Licks R Jordan and F Perez-Gonzalez ldquoAngleQIM a novel watermark embedding scheme robust againstamplitude scaling distortionsrdquo in Proceedings of the IEEE Inter-national Conference on Acoustics Speech and Signal Processing(ICASSP rsquo05) pp 797ndash800 March 2005
[26] H R Fazlali S Samavi N Karimi and S Shirani ldquoAdaptiveblind image watermarking using edge pixel concentrationrdquoMultimedia Tools and Applications vol 76 no 2 pp 3105ndash31202017
[27] X-Y Wang Y-N Liu H Xu A-L Wang and H-Y YangldquoBlind optimum detector for robust image watermarking in
16 Security and Communication Networks
nonsubsampled shearlet Domainrdquo Information Sciences vol372 pp 634ndash654 2016
[28] X Liu C Lin and S Yuan ldquoBlind Dual Watermarking forColor Imagesrsquo Authentication and Copyright Protectionrdquo IEEETransactions on Circuits and Systems for Video Technology vol28 no 5 pp 1047ndash1055 2018
[29] Poonam and S M Arora ldquoA DWT-SVD based Robust DigitalWatermarking for Digital Imagesrdquo Procedia Computer Sciencevol 132 pp 1441ndash1448 2018
[30] S Fazli and M Moeini ldquoA robust image watermarking methodbased on DWT DCT and SVD using a new technique forcorrection of main geometric attacksrdquo Optik - InternationalJournal for Light and ElectronOptics vol 127 no 2 pp 964ndash9722016
[31] K Kuppusamy and K Thamodaran ldquoOptimized Image Water-marking Scheme Based On PSOrdquo Procedia Engineering vol 38pp 493ndash503 2012
[32] V S Rao R S Shekhawat and V K Srivastava ldquoA dwt-dct-svd based digital image watermarking scheme using particleswarm optimizationrdquo in Proceedings of the 2012 IEEE StudentsrsquoConference on Electrical Electronics and Computer Science pp1ndash4 2012
[33] T T Takore P R Kumar and G L Devi ldquoA modified blindimage watermarking scheme based on DWT DCT and SVDdomain using GA to optimize robustnessrdquo in Proceedings ofthe 2016 International Conference on Electrical Electronics andOptimization Techniques ICEEOT 2016 pp 2725ndash2729 IndiaMarch 2016
[34] H Hu S Lin and L Hsu ldquoEffective blind speech watermarkingvia adaptive mean modulation and package synchronization inDWT domainrdquo EURASIP Journal on Audio Speech and MusicProcessing vol 2017 no 1 2017
[35] W-H Lin Y-R Wang S-J Horng T-W Kao and Y Pan ldquoAblind watermarking method using maximum wavelet coeffi-cient quantizationrdquo Expert Systems with Applications vol 36no 9 pp 11509ndash11516 2009
[36] X Li J Liu J Sun X Yang andW Liu ldquoStep-projection-basedspread transform dither modulationrdquo IET Information Securityvol 5 no 3 pp 170ndash180 2011
[37] M Hamghalam S Mirzakuchaki andM A Akhaee ldquoGeomet-ric modelling of the wavelet coefficients for image watermark-ing using optimum detectorrdquo IET Image Processing vol 8 no 3pp 162ndash172 2014
[38] V S Verma R K Jha and A Ojha ldquoDigital watermarkextraction using support vector machine with principal com-ponent analysis based feature reductionrdquo Journal of VisualCommunication and Image Representation vol 31 pp 75ndash852015
[39] C Li Z Zhang Y Wang B Ma and D Huang ldquoDithermodulation of significant amplitude difference for waveletbased robust watermarkingrdquo Neurocomputing vol 166 pp404ndash415 2015
Chemical EngineeringInternational Journal of Antennas and
Propagation
International Journal of
Hindawiwwwhindawicom Volume 2018
Hindawiwwwhindawicom Volume 2018
Navigation and Observation
International Journal of
Hindawi
wwwhindawicom Volume 2018
Advances in
Multimedia
Submit your manuscripts atwwwhindawicom
16 Security and Communication Networks
nonsubsampled shearlet Domainrdquo Information Sciences vol372 pp 634ndash654 2016
[28] X Liu C Lin and S Yuan ldquoBlind Dual Watermarking forColor Imagesrsquo Authentication and Copyright Protectionrdquo IEEETransactions on Circuits and Systems for Video Technology vol28 no 5 pp 1047ndash1055 2018
[29] Poonam and S M Arora ldquoA DWT-SVD based Robust DigitalWatermarking for Digital Imagesrdquo Procedia Computer Sciencevol 132 pp 1441ndash1448 2018
[30] S Fazli and M Moeini ldquoA robust image watermarking methodbased on DWT DCT and SVD using a new technique forcorrection of main geometric attacksrdquo Optik - InternationalJournal for Light and ElectronOptics vol 127 no 2 pp 964ndash9722016
[31] K Kuppusamy and K Thamodaran ldquoOptimized Image Water-marking Scheme Based On PSOrdquo Procedia Engineering vol 38pp 493ndash503 2012
[32] V S Rao R S Shekhawat and V K Srivastava ldquoA dwt-dct-svd based digital image watermarking scheme using particleswarm optimizationrdquo in Proceedings of the 2012 IEEE StudentsrsquoConference on Electrical Electronics and Computer Science pp1ndash4 2012
[33] T T Takore P R Kumar and G L Devi ldquoA modified blindimage watermarking scheme based on DWT DCT and SVDdomain using GA to optimize robustnessrdquo in Proceedings ofthe 2016 International Conference on Electrical Electronics andOptimization Techniques ICEEOT 2016 pp 2725ndash2729 IndiaMarch 2016
[34] H Hu S Lin and L Hsu ldquoEffective blind speech watermarkingvia adaptive mean modulation and package synchronization inDWT domainrdquo EURASIP Journal on Audio Speech and MusicProcessing vol 2017 no 1 2017
[35] W-H Lin Y-R Wang S-J Horng T-W Kao and Y Pan ldquoAblind watermarking method using maximum wavelet coeffi-cient quantizationrdquo Expert Systems with Applications vol 36no 9 pp 11509ndash11516 2009
[36] X Li J Liu J Sun X Yang andW Liu ldquoStep-projection-basedspread transform dither modulationrdquo IET Information Securityvol 5 no 3 pp 170ndash180 2011
[37] M Hamghalam S Mirzakuchaki andM A Akhaee ldquoGeomet-ric modelling of the wavelet coefficients for image watermark-ing using optimum detectorrdquo IET Image Processing vol 8 no 3pp 162ndash172 2014
[38] V S Verma R K Jha and A Ojha ldquoDigital watermarkextraction using support vector machine with principal com-ponent analysis based feature reductionrdquo Journal of VisualCommunication and Image Representation vol 31 pp 75ndash852015
[39] C Li Z Zhang Y Wang B Ma and D Huang ldquoDithermodulation of significant amplitude difference for waveletbased robust watermarkingrdquo Neurocomputing vol 166 pp404ndash415 2015