Top Banner
978-1-4673-6621-2/16/$31.00 © 2016 IEEE Discrete Wavelet Transform Based Video Watermarking Technique Sneha Kadu, VNIT, Nagpur, [email protected] Ch. Naveen, VNIT, Nagpur, [email protected] V. R. Satpute, VNIT, Nagpur, [email protected] A. G. Keskar, VNIT, Nagpur, [email protected] AbstractIn this paper, an effective algorithm for providing copyright protection is proposed by using a new embedding strategy for Discrete Wavelet Transform-based video watermarking. Discrete Wavelet Transform (DWT) is applied on the video, to convert the spatial data into frequency domain, having low pass and high pass components. The low frequency component is used for generating the key, by using the watermark image and the binarized Low frequency part (LL) of the video frame. Same procedure is applied on each frame to generate the key for corresponding frame. This generated key should be used at receiver for extracting the watermark which provides copyright protection. Blind watermarking technique is used in this paper which require only key to extract the embedded watermark. The original video is not required during extraction. To criticize the robustness of algorithm, the original watermark image is compared with extracted watermark image after several attacks and their Peak Signal to Noise Ratio (PSNR), Normalized Correlation Coefficient (NC) and Structural Similarity index (SSIM) are calculated. The experimental results demonstrate that the watermark is invisible and it is robust against the various attacks and addition of noise to the video. Keywords— Discrete Wavelet Transform (DWT); Digital Video Processing; Wavelets; Peak Signal to Noise Ratio (PSNR); Normalized Correlation Coefficient (NC); Structural Similarity index (SSIM) I. INTRODUCTION With a tremendous improvement in the fields of science and technology, there has been drastic growth in the Internet [2]. With the increase in demand of internet, broadband communication has also taken its own pace. This leads to the digital transfer of data such as images, videos, etc. [8]. Hence, it is important to avoid the unwanted redistribution of such data or unauthorized access by the illegal users [5, 6]. Digital Watermarking is one of the appealing methods to protect the copyright and unauthorized access to the content. Such type of watermarking can be applied to images, audio, videos etc. These watermarks should not alter the quality of content and it should be robust to the various attacks and distortion. Lot of work has been done on image watermarking and few has been done on video. This paper presents the digital watermarking applied on videos. The digital video watermarking is used to protect the video from digital manipulation and provides the copyright authentication [10, 11]. The two video watermarking methodologies to embed the watermark bit are Spatial Domain Watermarking and Spectral Domain Watermarking. Spatial Domain method is not robust to many signal processing attacks. Spectral Domain method ensures the robustness of watermark. Frequently used transforms are the Discrete Fourier Transform (DFT), Discrete Cosine Transform (DCT) and Discrete Wavelet Transform (DWT). The Discrete wavelet transform is more frequently use because of its excellent multi-resolution and spatial localization characteristics [9]. II. EXISTING WORK Gu Tiaming and Wang Yangie [1] proposed the robust image watermarking algorithm. This algorithm is based on DWT coefficient. The level-3 DWT is done to obtain the low frequency part and the information in this part is used along with watermark to obtain the key. The key generated is used to extract the watermark. This algorithm does not change the content of original image and thus robust to various attacks. Hong, Kim and Han [2] proposed a watermarking scheme in which watermark is embedded in middle frequency band of the two level DWT. They have generated the pseudo random code which is used for watermark positioning in the LH2 (level-2 LH component) band. The mean value of neighboring pixel is found out and this mean value is then compared with the selected pixel. Replacement of selected pixel is done based on the comparison and new flags are generated consisting of values 0 and 1 based on the conditions suggested by the authors. This generated flag is stored and used for watermark extraction. Daren, Jiufen, Jiwu and Hongmie [3] embed watermark in low frequency sub-band first and remaining in higher frequency sub bands depending on the significance of sub band. Watermark embedding is done with different embedding formulae. This algorithm incorporates the features of visual masking of human vision system in watermarking III. PROPOSED VIDEO WATERMARKING ALGORITHM Wavelets are the Basal functions that are used for representing signals. These functions are analogous to sine and cosines in Fourier analysis. The Discrete Wavelet Transform (DWT) yields the fast computation of Wavelet transform. In this paper, DWT is applied on each frame and the video watermarking is done on every frame to which DWT is applied. For 2-D frame, applying DWT means applying 1-D filter in two dimensions. The filter then divide the frame into four non overlapping sub-bands called as LL1, LH1, HL1 and HH1. In this paper, L stands for low pass, H stands for high pass, while, the number indicates the level of DWT applied. To obtain the next level, the LL1 sub-band is selected and
6

Discrete Wavelet Transform Based Video Watermarking Technique wavelet... · Discrete Wavelet Transform Based Video ... strategy for Discrete Wavelet Transform-based video watermarking.

Apr 24, 2018

Download

Documents

hoangduong
Welcome message from author
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
Page 1: Discrete Wavelet Transform Based Video Watermarking Technique wavelet... · Discrete Wavelet Transform Based Video ... strategy for Discrete Wavelet Transform-based video watermarking.

978-1-4673-6621-2/16/$31.00 © 2016 IEEE

Discrete Wavelet Transform Based Video Watermarking Technique

Sneha Kadu, VNIT, Nagpur, [email protected]

Ch. Naveen, VNIT, Nagpur, [email protected]

V. R. Satpute, VNIT, Nagpur, [email protected]

A. G. Keskar, VNIT, Nagpur, [email protected]

Abstract— In this paper, an effective algorithm for providing

copyright protection is proposed by using a new embedding strategy for Discrete Wavelet Transform-based video watermarking. Discrete Wavelet Transform (DWT) is applied on the video, to convert the spatial data into frequency domain, having low pass and high pass components. The low frequency component is used for generating the key, by using the watermark image and the binarized Low frequency part (LL) of the video frame. Same procedure is applied on each frame to generate the key for corresponding frame. This generated key should be used at receiver for extracting the watermark which provides copyright protection. Blind watermarking technique is used in this paper which require only key to extract the embedded watermark. The original video is not required during extraction. To criticize the robustness of algorithm, the original watermark image is compared with extracted watermark image after several attacks and their Peak Signal to Noise Ratio (PSNR), Normalized Correlation Coefficient (NC) and Structural Similarity index (SSIM) are calculated. The experimental results demonstrate that the watermark is invisible and it is robust against the various attacks and addition of noise to the video.

Keywords— Discrete Wavelet Transform (DWT); Digital Video Processing; Wavelets; Peak Signal to Noise Ratio (PSNR); Normalized Correlation Coefficient (NC); Structural Similarity index (SSIM)

I. INTRODUCTION With a tremendous improvement in the fields of science

and technology, there has been drastic growth in the Internet [2]. With the increase in demand of internet, broadband communication has also taken its own pace. This leads to the digital transfer of data such as images, videos, etc. [8]. Hence, it is important to avoid the unwanted redistribution of such data or unauthorized access by the illegal users [5, 6]. Digital Watermarking is one of the appealing methods to protect the copyright and unauthorized access to the content. Such type of watermarking can be applied to images, audio, videos etc. These watermarks should not alter the quality of content and it should be robust to the various attacks and distortion.

Lot of work has been done on image watermarking and few has been done on video. This paper presents the digital watermarking applied on videos. The digital video watermarking is used to protect the video from digital manipulation and provides the copyright authentication [10, 11]. The two video watermarking methodologies to embed the watermark bit are Spatial Domain Watermarking and Spectral Domain Watermarking. Spatial Domain method is not robust to many signal processing attacks. Spectral Domain method

ensures the robustness of watermark. Frequently used transforms are the Discrete Fourier Transform (DFT), Discrete Cosine Transform (DCT) and Discrete Wavelet Transform (DWT). The Discrete wavelet transform is more frequently use because of its excellent multi-resolution and spatial localization characteristics [9].

II. EXISTING WORK Gu Tiaming and Wang Yangie [1] proposed the robust

image watermarking algorithm. This algorithm is based on DWT coefficient. The level-3 DWT is done to obtain the low frequency part and the information in this part is used along with watermark to obtain the key. The key generated is used to extract the watermark. This algorithm does not change the content of original image and thus robust to various attacks.

Hong, Kim and Han [2] proposed a watermarking scheme in which watermark is embedded in middle frequency band of the two level DWT. They have generated the pseudo random code which is used for watermark positioning in the LH2 (level-2 LH component) band. The mean value of neighboring pixel is found out and this mean value is then compared with the selected pixel. Replacement of selected pixel is done based on the comparison and new flags are generated consisting of values 0 and 1 based on the conditions suggested by the authors. This generated flag is stored and used for watermark extraction.

Daren, Jiufen, Jiwu and Hongmie [3] embed watermark in low frequency sub-band first and remaining in higher frequency sub bands depending on the significance of sub band. Watermark embedding is done with different embedding formulae. This algorithm incorporates the features of visual masking of human vision system in watermarking

III. PROPOSED VIDEO WATERMARKING ALGORITHM Wavelets are the Basal functions that are used for representing signals. These functions are analogous to sine and cosines in Fourier analysis. The Discrete Wavelet Transform (DWT) yields the fast computation of Wavelet transform. In this paper, DWT is applied on each frame and the video watermarking is done on every frame to which DWT is applied. For 2-D frame, applying DWT means applying 1-D filter in two dimensions. The filter then divide the frame into four non overlapping sub-bands called as LL1, LH1, HL1 and HH1. In this paper, L stands for low pass, H stands for high pass, while, the number indicates the level of DWT applied. To obtain the next level, the LL1 sub-band is selected and

Page 2: Discrete Wavelet Transform Based Video Watermarking Technique wavelet... · Discrete Wavelet Transform Based Video ... strategy for Discrete Wavelet Transform-based video watermarking.

again it is divided into four non overlappingLL2, LH2, HL2 and HH2 [4,7]. The process DWT is shown in Fig. 1(a). The Fig. 1(spatial multi-resolution analysis (MRA) appsuch as images or frames.

The low pass component has higher pand includes most of the energy of originasub-band is also termed as the low pass apporiginal image. Moreover, watermark evarious attacks like data compression, losubsampling D/A and A/D conversions, cancontent of low pass sub-bands compared to In the proposed algorithm, level-3 DWT is cthe watermark embedding in LL3 i.e. low frof level-3 DWT is performed on video frame

(a)

Fig. 1. (a) The process of spatial 2D- DWT; (b) Spatial

A. Key Generation Procedure Using WatermAnd LL Part The watermarking embedding procedure

be described in the following steps.

Step 1: Apply DWT on each frame toriginal video frames into various sub-bandwavelet decomposition. This leads to frequen

(a)

Fig. 2. For ‘viptraffic.avi’ video frame #20; (a) (b) Level-3 2-D DWT

g sub-bands named of computing 2-D

(b) represents the plied on 2-D data

erceptual capacity al image [3]. This proximation of the encountering with ow pass filtering, n change very less

higher sub-bands. computed and then requency sub-band s. [5]

(b) 2D- DWT with MRA

marked Image

for each frame can

to decompose the s using 2-D DWT

ncy band LL3.

(b)

original frame

(a)

Fig. 3. For ‘Video_1.avi’ video(b) Level-3

(a)

Fig. 4. For ‘Video_4.avi’ vide(b) Level-3

Step 2: Obtain the me

coefficient using formula, giframe.

∑=

=N

iia

NT kk

1)(1

Where P is the total numgiven video, kT represents a vfor the frames of the videocoefficients i.e. LL3 componenN represents the total number component of corresponding fra

Step 3: Compare the lowmean value kT for kth frame ob

10Where, k = 1, 2, 3,…,P whichleads to generating a binary importance of DWT coefficiethreshold obtained. This can coefficients for the generation o

Step 4: Generate the key and for kth frame.

Where is the watermark imextract this watermark during th

For example, the matrix 1; 1 1 1 1 0 0 1 1] of dimensbe also of same dimension and 0 0 0 0 1 01]. The key is operation on and .

=[1 0 0 1 1 0 0 1;1 1 1 1 0 0 1 0 1]

Therefore, key matrix is [1 1

(b)

o frame #410; (a) original frame 3 2-D DWT

(b)

eo frame #47; (a) original frame 3 2-D DWT

ean value of low frequency iven in equation (1) for each

k = 1, 2,...,P (1)

mber of frames available in the vector consisting of P thresholds o, is low frequency DWT nts of the original video frames, of coefficients available in LL3 ame.

w frequency coefficient with the tained in step 2.

(2a) (2b)

h indicates frame number. This y pattern corresponding to the ents in LL3 part based on the

be termed as weights of the of key.

by applying XOR operation on

⊗ (3) mage, and, Kk is the key used to

he watermark recovery process.

is taken to be [1 0 0 1 1 0 0 ions 2 8 .The matrix will is taken to be [0 1 1 1 0 1 0 0; 1 generated by applying Xoring

1 1] [0 1 1 1 0 1 0 0;1 0 0 0 0

1 1 0 11 1 0;0 1 1 1 0 1 1 0].

Page 3: Discrete Wavelet Transform Based Video Watermarking Technique wavelet... · Discrete Wavelet Transform Based Video ... strategy for Discrete Wavelet Transform-based video watermarking.

This key need to be known to the authorized personnel.

Step 5: Apply IDWT to the video obtained from step 4 to obtain the watermarked video.

Fig. 5. Watermark embedding process Fig. 6. Watermark extraction

process

B. Watermark Extraction Procedure Using Key The advantage of this algorithm is that one need to know only the keys i.e. only is required to obtain the watermark of each frame. The extraction procedure can be described in the following steps.

Step 1: Apply DWT on each frame to decompose the watermarked video frames into various sub-bands using 2-D DWT wavelet decomposition. This leads to frequency band LL3.

Step 2: Obtain the mean value of low frequency coefficients using the formula as given in equation (4) for each video frame.

∑=

=N

iia

NT kk

1)('1' k = 1, 2,...,P (4)

Where P is the total number of frames available in the watermarked video, kT ' represents a vector consisting of P thresholds for the watermarked frames of the video, is low frequency DWT coefficients i.e. LL3 components of the watermarked video frames, N represents the total number of coefficients available in the LL3 component of the corresponding watermarked frame. Step 3: Compare the low frequency coefficients with the mean value for kth frame obtained in step 2. 1 (5a)0 (5b)Where, k = 1, 2, 3,…,P which indicates frame number. This leads to generating a binary pattern corresponding to the importance of DWT coefficients in LL3 part based on the threshold obtained. This can be termed as weights of the coefficients for the extraction of watermark. Step 4: Generate watermark by applying XOR operation on and .

⊗ (6) Where the extracted watermark image, and, is the key obtained for each frame during the watermark embedding process.

Step 5: Apply IDWT to the video obtained from step 4 to get back the original video after extraction of watermark.

IV. EXPERIMENTAL RESULTS The proposed Algorithm is implemented using MATLAB as the basic tool on the video database consisting of videos downloaded from internet, standard video databases and also on the videos captured in the laboratory under non-standard conditions as specified in Table 1. The computations were carried out on the system with configuration details as, Intel(R) Core™ i7-4770 processor having maximum clock frequency of 3.4 GHz and 32 GB of RAM.

(a) (b)

(c) (d)

Fig. 7. For ‘viptraffic.avi’; (a) original frame #20 (b) original watermark testpat1, (c) watermarked video frame, (d) Extracted water mark from (c)

(a) (b)

(c) (d)

Fig. 8. For Video_1.avi’; (a) original frame #410 (b) original watermark testpat1, (c) watermarked video frame, (d) Extracted water mark from (c)

Page 4: Discrete Wavelet Transform Based Video Watermarking Technique wavelet... · Discrete Wavelet Transform Based Video ... strategy for Discrete Wavelet Transform-based video watermarking.

(a) (b)

(c) (d)

Fig. 9. For ‘Video_4; (a) original frame #47 (b) original watermark testpat1, (c) watermarked video frame, (d) Extracted water mark from (c)

The algorithm is tested on a large number of videos but here, in this paper only ten videos are indicated. The figures are shown only for one video i.e. viptraffic.avi which is a standard video available in the MATLAB directory. The watermark image used in this work is again taken from standard MATLAB database i.e. testpat1which is having dimensions of 256×256 which is resized then to the size of the LL3 part of the DWT decomposition applied on the frames. Results of no attack are shown in fig. 7 to Fig. 9. Fig. 7 indicates the results of proposed work for frame number 20 of the viptraffic video. Fig. 8 indicates the frame number 410 of Video_4 (Standard database) and Fig. 9 indicate results for Video_4(Captured in lab) for frame number 47.

The watermarked video quality is measured using PSNR (Peak Signal to Noise Ratio). For calculation of PSNR, one needs to obtain the mean squared error (MSE). The equation for MSE is given in equation (7) as follows: MSE L ∑ ∑ f i, j f i, jWH (7)

Where is the original frame, 'f is the watermarked frame, and L=W×H i.e. it gives the total number of pixels in the video frame. PSNR in decibel (dB) for individual video frame is obtained by using equation (8) as follows: PSNR 10log RMSE ; k 1, 2, 3, … . . , P (8) Where, R is the maximum possible value in the corresponding frame and P indicates total number of frames. For video, PSNR is computed here by taking average of PSNR values of corresponding frames of the video. This average value is computed using the equation (9) as follows: PSNRV P ∑ PSNRP (9) Where, the average value of the PSNR for video and P is the total number of frames of the corresponding video. The next parameter to be computed is Normalized Correlation (NC) is used to show the similarities between the original watermark and the extracted watermark. The value of NC if 1, then it indicates that watermarked frame is exactly

matching the original frame. On the contrast, if the value of the NC is 0, it indicates that the frames are totally different and hence not matching. Otherwise, the NC value lies in between 0 and 1. The more the value of NC, the more closer we are towards the original image. For each frame in the video, the value of NC is obtained by using the formula as given in equation (10). Here the Exclusive-NOR operation on the original watermark and extracted watermark is performed to get the NC value. The bar indicates the ex-NOR operation. NC N ∑ ∑ g i, j g i, jYX ; k 1, 2, … . , P (10)

Where, N=X×Y, which represents the total number of pixels of watermark image for corresponding frame.

Thus, for a video, the NC value is computed as average of the NC of each frame using equation (11).

NCV 1P NC 11P

The next parameter to be computed is SSIM (Structural Similarity) index. This parameter is used to measure the similarity between two images [12]. SSIM x, y ∑ SSIM x, y, kP P ; k 0, 1, … , P 12 Where, P is the number of frames of given video. The term SSIM(x, y, k) is defined as, SSIM x, y, k 2µ µ c 2σ cµ µ c σ σ c 13 Where, ‘x’ is the original watermark and ‘y’ is the extracted watermark from video frames, is the mean of the intensities available in the original watermark from xth frame, is the mean of the intensities available in the extracted watermark from yth frame, is the variance of original watermark i.e ‘x’ from xth frame, is the variance of extracted watermark from yth frame, and is covariance of original and extracted watermark i.e. ‘x’ and ‘y’.

Fig. 10. NC of watermark (No attack)

The graphs shown in Fig. 10, represents the Normalized Correlation (NC) of the watermark when no attack is applied on the video.

To investigate the robustness of the proposed watermarking algorithm, the watermarked video is attacked using average filter, median filter, Gaussian noise, salt and pepper noise and histogram equalization. The results of which are indicated in

Page 5: Discrete Wavelet Transform Based Video Watermarking Technique wavelet... · Discrete Wavelet Transform Based Video ... strategy for Discrete Wavelet Transform-based video watermarking.

figure 11 and 15 and the parameter are tabulated in Table 2.The received watermark after applying various attacks are also shown in Table 3. The results summarized below for various operations shows that the proposed scheme is robust and secure against various processing operation such as averaging, noise addition, histogram equalization.

Fig 11 and Fig 12 shows the Normalized Correlation of watermark after average filtering and median filtering. The average Normalized Correlation of watermark after applying average filter and median filter on watermarked video is found to be 0.993 and 0.987 respectively. This shows the robustness of average and median attack on watermark image. Fig 13 and Fig 14 shows the robustness of watermark on each frame after the addition of Gaussian and salt and pepper noise respectively. The average normalized coefficient for noises is 0.956 and 0.947 respectively. Fig 15 shows the robustness against Histogram Equalization and its Normalized correlation for video is found to be around 0.969.

The Normalized Correlation for each attack is calculated and it is found that it is approaching towards 1. It means that extracted watermark is much similar to the original watermark

V. CONCLUSION The robust video watermarking algorithm is proposed by

embedding watermark on each frame of the video. This algorithm realizes blind watermarking with watermark detection and extraction and is found to be robust to most common attacks. It is also observed from Table 2 and 3 that the proposed method works well for the watermarking of the video contents. The Normalized Correlation Coefficient (NC) and Structural Similarity (SSIM) index are approaching towards the value 1 which indicates that reconstructed watermark is matching to that of the original one.

REFERENCES [1] Gu Tianming and Wang Yanjie, “DWT-based Digital image

Watermarking Algorithm,” 2011 10th International Conference on (Volume:3 ),pp.163-166, Aug 2011.

[2] Ikpyo Hong, Intaek Kim, and Seung-Soo Han “A Blind watermarking technique using Wavelet transform,” 2001. Proceedings. ISIE 2001. IEEE International Symposium on Industrial Electronics vol.3 ,pp.1946-1950,June 2001.

[3] Huang Daren, Liu Jiufen, Huang Jiwu, Liu Hongmei, “A DWT based Image Watermarking Algorithm,” 2001 IEEE international conference on Multimedia and expo, , pp. 429-432,Aug 2001.

[4] Xiaonion Tang, Quan Wen ,Guijin Nian, Jianchun Wang and Huiming Zhu,” An improved Robust watermarking technique in Wavelet Domain,” 2010 Second International Conference on Multimedia and expo,Vol :1, pp.270-273, April 2010..

[5] Y. Raghavender Rao, Dr.E.Nagabhooshanam, Nikhil Prathapani, “Robust video watermarking Algorithm Based on SVD transform”, 2014 International Conference on Information Communication and Embedded System ICICES2014, pp.1-5, Feb 2014.

[6] Quin Liu and Jun Ying, “Grayscale image Digital Watermarkinmg Technology Based on Wavelet Analysis”, 2012 IEEE symposium on Electrical and Electronics System(EEESYM), pp. 618-621 ,June 2012.

[7] Hyung Kyo Lee ,Hee Jung Kim,Ki-Ryong Kwon ,Jon-Keuk Lee, “ROI Medical Image Watermarking using DWT and Bit plane”, 2005 Asia-Pacific Conference on Communications, Perth, Western Australia, pp. 512-515 ,3- 5 October 2005.

[8] G.Ramkumar and M.Manikandan, “Uncompressed Digital Video Watermarking Using Stationary Wavelet Transform,” 2014 IEEE International Conference on Advanced Communication Control and Computing Technologies (ICACCCT), pp.1252-1258, May 2014.

[9] Pragya Agarwal, and Ankur Choudhary , “Protecting Video Data Through Watermarking: A Comprehensive Study,” 2014 5th International Conference- Confluence The Next Generation Information Technology Summit(Confluence),pp.657-662,Sept 2014.

[10] Bhavna Goel and Charu Agarwal, “An Optimized Un-compressed Video Watermarking Scheme based on SVD and DWT ” 2013 Sixth International Conference on Contemporary Computing (IC3), pp. 307-312, Aug 2013.

[11] Mr Mohan A Chimanna and Prof.S.R.Khot , “Robustness of video watermarking against various attacks using wavelet Transform Techniques and Principle Component Analysis”, 2013 International Conference on Information Communication and Embedded System(ICICES),pp.613,Feb 2013.

[12] Zhou Wang, Alan C. Bovik, Hamid R. Sheikh, Eero P. Simoncelli, “Image Quality Assessment: From Error Visibility to Structural Similarity” IEEE Transactions on Image Processing Vol 13,pp.1-4,April 2004.

Fig.11:NC for Average Filter Fig.12: NC for Median Filter Fig.13: NC for Gaussian Noise

Fig.14: NC for Salt and Pepper Noise Fig.15: NC for Histogram Equalization

Page 6: Discrete Wavelet Transform Based Video Watermarking Technique wavelet... · Discrete Wavelet Transform Based Video ... strategy for Discrete Wavelet Transform-based video watermarking.

Table 1. Video database

Sr. No. Video Name Specifications

Source Frame size Durtion(in sec)

No. of frames

1 viptraffic.avi 120×160 8 120 Standard Database 2 rhinos.avi 240×320 6 114 Standard Database 3 Shaky_car.avi 240×320 4 132 Standard Database 4 Scenevideoclip.avi 160×120 7 92 Standard Database 5 Video_1.avi 240×320 44 448 Downloaded 6 Video _2.avi 288×360 43 1077 Downloaded 7 Video_3.avi 480×640 5 67 Downloaded 8 Video_4.avi 640×480 18 543 Captured in lab 9 Video_5.avi 480×640 14 140 Captured in lab

10 Video_6.avi 480×640 15 155 Captured in lab

Table2.Experimental results of watermark after various attacks on watermarked video Video Parameters No

attack Average filter

Median filter

Gaussian Noise

Salt and Pepper Noise

Histogram equalization

viptraffi.avi PSNR(dB) Inf 22.101 19.749 13.778 12.93 15.970 NC 1 0.993 0.987 0.956 0.947 0.969 SSIM 1 0.994 0.989 0.957 0.948 0.969

rhinos. avi

PSNR(dB) Inf 26.637 25.921 17.7435 17.165 12.57 NC 1 0.997 0.997 0.9776 0.975 0.9301 SSIM 1 0.998 0.998 0.9842 0.983 0.9258

scenevideoclip.avi

PSNR(dB) Inf 21.714 19.094 18.883 18.020 11.4421 NC 1 0.993 0.985 0.986 0.982 0.896 SSIM 1 0.993 0.986 0.987 0.986 0.929

shaky_car.avi PSNR(dB) Inf 25.914 24.817 17.864 16.876 7.934 NC 1 0.997 0.996 0.983 0.979 0.838 SSIM 1 0.998 0.997 0.992 0.989 0.934

Video_1.avi(net) PSNR(dB) Inf 27.842 28.759 19.946 19.183 13.685 NC 1 0.998 0.998 0.989 0.987 0.942 SSIM 1 0.999 0.999 0.993 0.992 0.942

Video_2.avi(net) PSNR(dB) Inf 25.428 22.071 16.076 15.351 17.964 NC 1 0.997 0.993 0.975 0.975 0.983 SSIM 1 0.998 0.995 0.981 0.977 0.987

Video_3.avi(net) PSNR(dB) Inf 27.908 25.682 18.071 17.120 13.320 NC 1 0.998 0.997 0.984 0.997 0.953 SSIM 1 0.998 0.998 0.988 0.986 0.967

Video_4.avi(lab1)

PSNR(dB) Inf 32.350 32.453 19.919 19.633 8.893 NC 1 0.999 0.999 0.989 0.989 0.870 SSIM 1 0.999 0.999 0.992 0.992 0.899

Video_5.avi(lab2)

PSNR(dB) Inf 27.716 27.674 15.649 14.9470 7.575 NC 1 0.998 0.998 0.972 0.967 0.824 SSIM 1 0.998 0.998 0.981 0.977 0.880

Video_6.avi(lab3)

PSNR(dB) Inf 25.143 24.959 15.632 14.973 11.221 NC 1 0.997 0.997 0.972 0.968 0.924 SSIM 1 0.998 0.997 0.981 0.978 0.948

Table3. Extracted watermark of frame#20 after various attacks on viptraffic video

Video Name Parameter No attack Average

Filter Median Filter

Gaussian Filter

Salt and Pepper

Histogram Equalization

viptraffic.avi

Extracted Watermark