Decade progress of palmprint recognition: A brief surveyancai/DIP/articole/Decade progress of... · [3], palm vein [4] , palmprint [5] and other features [6] in personal verification
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
Neurocomputing 328 (2019) 16–28
Contents lists available at ScienceDirect
Neurocomputing
journal homepage: www.elsevier.com/locate/neucom
Decade progress of palmprint recognition: A brief survey
Dexing Zhong
∗, Xuefeng Du, Kuncai Zhong
School of Electronic and Information Engineering, Xi’an Jiaotong University, Xi’an 710049, China
a r t i c l e i n f o
Article history:
Received 4 November 2017
Revised 14 February 2018
Accepted 22 March 2018
Available online 20 August 2018
Keywords:
Palmprint recognition
ROI (region of interest)
Feature extraction
Matching
Fusion
a b s t r a c t
As an advanced research topic in biometrics techniques, palmprint recognition has been fully studied for
more than 20 years. Due to its superiority to other biological features, i.e. high recognition accuracy and
convenience for practical application, many research achievements on palmprint have emerged recently,
especially in the past decade. This paper presents a comprehensive overview of recent research progress
of palmprint recognition as well as the basic background knowledge for it. In addition, it mainly focuses
on data acquisition, database, preprocessing, feature extraction, matching and fusion. Ultimately, we dis-
cuss the challenges and future perspectives in palmprint recognition for further works.
D. Zhong et al. / Neurocomputing 328 (2019) 16–28 17
Fig. 1. Palmprint features in (a) a high-resolution image and (b) a low-resolution
image [13] .
u
t
h
s
t
p
b
d
n
i
i
G
m
m
d
C
t
s
a
s
l
c
d
d
q
g
u
w
k
l
m
i
w
m
p
S
i
r
o
t
t
d
2
2
b
D
t
s
f
f
t
o
e
s
w
d
d
d
a
p
a
F
d
t
p
[
d
m
s
p
b
s
m
2
m
t
d
s
e
i
t
l
p
e
p
s
a
In terms of multispectral methods, they use features extracted
nder distinctive spectral wavelengths for identification to improve
he accuracy and anti-spoof capability [5] . Guo et al. [19] analyzed
yperspectral palmprint data to determine the optimal number of
pectral bands and obtained the most typical bands to establish
heir recognition system.
So far, 3D palmprint recognition has undergone essential
rogress as well. While 2D images can be counterfeited by law-
reakers or contaminated by noise [20] , 3D images contain more
epth information and are gradually used to improve the robust-
ess of recognition. Zhang et al. [21] exploited the 3D structural
nformation of the palm and proposed the structured light imag-
ng to establish palmprint datasets. Then they extracted mean and
aussian curvature image, surface type for classification that are
ore stable to illumination variations.
In addition, some researchers found that using multimodal bio-
etrics could significantly improve the recognition rate because
ifferent features serve as mutual supplement to each other. Thus,
ui et al. [22] applied principal component analysis (PCA) and
wo-phase test sample representation (TPTSR) to present the fu-
ion scheme of 2D and 3D features. Face and palmprint features
re fused by SDA-GSVD (subclass discriminant analysis-generalized
ingular value decomposition) [23] , which outperforms some re-
ated multimodal recognition methods.
To our best knowledge, a classic palmprint recognition pro-
ess is composed of five sections: palmprint image acquisition,
atabase, preprocessing, feature extraction and matching that are
emonstrated in Fig. 2 .
The acquisition device obtains palmprint images of different
ualities that are consistent with subsequent recognition. The re-
ion of interest (ROI) is the core of preprocessing stage. Commonly
sed algorithms are the reference coordinate system method,
hich is showed in Fig. 3 . In view of the feature extraction, several
inds of algorithms were proposed [5] , such as subspace methods,
earning methods, line-based and coding-based approaches. Each
ethod extracts features from a global or local scope, and each has
ts own advantages. The matching process matches testing samples
ith other samples in the database based on a certain predeter-
ined matcher.
The rest of the paper focuses on the development of palm-
rint recognition in the recent decade within five sections.
ection 2 sums up emerging acquisition devices and correspond-
ng datasets. It also retrospects distinctive preprocessing algo-
ithms. Section 3 explains most of novel feature extraction meth-
ds while demonstrating matching means. Section 4 focuses on
he palmprint-related fusion. Section 5 discusses several summa-
ion points and offers future directions for further development in
etail.
. Image acquisition and preprocessing
.1. Image acquisition
When it comes to the acquisition algorithms, each of them is
ased on a specific database and a concrete application direction.
ue to the changeable environment in the real world, many effec-
ive algorithms proposed in the ideal acquisition condition are not
uitable for practical application of palmprint recognition. There-
ore, it is important to set up different databases to simulate dif-
erent conditions and test whether a particular algorithm fits into
he research environment. Then some modifications can be carried
ut to achieve a better experimental result.
In the recent decade, great deals of new databases have been
stablished, such as the examples in [4,12,16,24–31] . Except that
ome databases use traditional cameras and classic acquisition
ays [25,29,31] , i.e. CCD-based (charge-coupled device) scanners,
igital cameras, video cameras, to collect palmprint images, many
atabases are built adopting new devices that capture images on
ifferent platforms [4,27,32] . For instance, Aykut et al. [33] used
CCD camera, direct current (DC) auto iris lens, hand placement
latform and uniform LED (light emitting diode) light sources to
ccomplish online palm image acquisition, which was revealed in
ig. 4 .
In all image types, 2D palmprint data is the most widely used
ata because it is easily accessible and can be handled easily
oo. Meanwhile, there are also many databases containing other
almprint information, such as 3D [12,21,24,34,35] , multispectral
19,26,30] , and minutiae [25] . All three sorts of these images are
emonstrated in Figs. 5 and 6 .
To describe the developmental process of acquisition techniques
ore clearly, we list representative results in Table 1 , showing the
ummary of different databases established in recent decade. Ap-
roximately 25 new palmprint databases have been established,
ringing the total number of samplers to 4,200. It is sufficient to
imulate real and different conditions, which is a basic require-
ent in image obtaining process.
.2. Preprocessing
Preprocessing is the foundation for feature extraction and
atching. The quality of preprocessing has a significant impact on
he outcomes of recognition. In this paper, we mainly focus on the
evelopment of algorithms to extract ROI. Because it is the major
tep in preprocessing stage apart from other procedures like image
nhancement, image filtering and so on.
In the recent decade, distance is the most momentous measur-
ng target in ROI extraction. This method keeps a fixed pixel dis-
ance between the edge of ROI and the connection line of the val-
ey points [4,21,27,44,45] . However, because of the size variety of
almprint images, valuable area for feature extraction will not be
xtracted accurately if researchers only apply the distance princi-
le. Thus, the result of recognition cannot reach the highly desired
tandard. Consequently, other measures are employed lately, such
s ratio [19,46] and angle [47,48] .
18 D. Zhong et al. / Neurocomputing 328 (2019) 16–28
Fig. 2. The flow-process diagram of palmprint recognition system.
Fig. 3. Classic steps of preprocessing: (a) original image, (b) binary image, (c) boundary tracking, (d) building a coordinate system, (e) extracting the central part and (f) ROI
sample [19] .
Fig. 4. External view of palm image acquisition system and a hand placed to the platform [33] .
a
v
t
t
e
d
m
d
The principle of ratio is that the size of ROI accounts for a fixed
ratio in palmprint images. Angle principle makes use of the princi-
ple that ROI boundary point-valley point connection line and val-
ley point connection line have a constant angle. And according to
many experiments, the angle selected to be 45 ° or 60 ° is suitable
for precise feature extraction [47] . Paradigms of three distinctive
methods are illustrated in Figs. 7 –9 . These above approaches can
greatly decrease the error rate caused by variance of image size,
rotation and other environmental defects. Due to occurrence of
dverse factors in an image, like overlapping, different number of
alley points will result in different consequences of ROI extrac-
ion. Most of articles used 2 ∼6 valley points [4,19,21,27,44,45] and
here were also some using 12 [49] and 15 [33] valley points. Av-
rage number is five to our best knowledge.
Table 2 shows the summary of ROI extraction algorithms. A ten-
ency can be discovered from the chart, ratio and distance were
ore widely used nowadays [43,46,49] , and the fusion of ratio and
istance were applied more often in recent 2 years [16,51] .
D. Zhong et al. / Neurocomputing 328 (2019) 16–28 19
Fig. 5. (a) 2D palmprint image, (b) Extracted ROI of 2D image [36] and the minutiae of high resolution image [25] . .
Fig. 6. The above row shows the 3D palmprint images. The below row shows the 2D palmprint images [37] .
Table 1
Summary of different databases established in the recent decade. Device expresses whether the re-
search uses new device or platform (Y is yes and N is not). Data is the kind of information obtained
from the database. Number refers to the number of pictures that belong to one people in the dataset.
Device(Y/N) Data(2D/3D/Multispectral/Minutiae) Number Year Article
N 2D 40 2008 [29]
N 2D 346 2008 [29]
Y 2D 120 2008 [4]
N 2D 146 2009 [38]
Y 2D 150 2009 [32]
Y 3D 260 2009 [21]
N Minutiae \ 2011 [25]
Y Multispectral \ 2011 [26]
N 2D \ 2011 [28]
Y Multispectral 190 2012 [19]
Y 2D 100 2012 [27]
Y 2D 193 2012 [39]
Y Multispectral 500 2012 [30]
N 2D 500 2013 [31]
Y 2D 40 2013 [33]
Y 2D 20 2013 [40]
Y 3D 200 2013 [35]
N 2D 100 2014 [41]
N 2D 60 2015 [42]
N 2D 60 2015 [42]
Y 3D 100 2015 [34]
N 2D 75 2015 [43]
Y 3D 138 2017 [12]
N 2D & 3D 260 2017 [24]
Y 2D 600 2017 [16]
Fig. 7. Five steps of preprocessing based on the distance [50] .
20 D. Zhong et al. / Neurocomputing 328 (2019) 16–28
Fig. 8. ROI location using the ratio method: (a) palm width L determination, (b) ROI
creation with [OO1] = 1/10 L and [E1E2] = 2/3 L [46] .
Fig. 9. ROI extraction using the angle method: (a) Scanned image. (b) Binarized
image. (c) Hand contour and reference points. (d) Relevant points and palmprint.
(e) Palmprint region in gray scale hand image. (f) Extracted palmprint [47] . .
Table 2
Summary of ROI extraction. Ratio, angle and distance express the prin-
ciple of ROI extraction. The number in the second line denotes the
number of valley points utilized.
Ratio/Angle/Distance Valley Points Year Article
Distance 2 2008 [4]
Distance 2 2009 [45]
Distance 6 2009 [21]
Distance 2 2011 [44]
Ratio 2 2012 [19]
Distance 3 2012 [27]
Angle 6 2012 [47]
Angle 6 2013 [48]
Ratio 4 2014 [46]
Distance 12 2014 [49]
Distance 4 2015 [43]
Ratio & Distance 5 2016 [51]
Ratio & Distance 4 2017 [16]
t
a
G
a
t
i
o
3
t
e
i
[
r
w
a
i
s
(
a
d
m
fi
s
a
d
O
m
p
a
d
G
p
e
a
o
a
c
I
v
b
m
n
c
u
I
t
[
(
n
k
t
p
r
a
a
[
i
t
3. Feature extraction and matching
3.1. Feature extraction
Feature is the main index for comparison. The work of feature
extraction is for the sake of maximizing difference of different peo-
ple and similarity of the same ones. The algorithms to extract fea-
tures have developed rapidly in recent decade, not only about the
types of feature, but also about ways to record feature accurately
and effectively.
3.1.1. Discussion about the extracted features
The principal features in the recent decade are texture
Many researchers applied machine learning [75–81] either for
feature extraction or for classification. In the latest 3 years, because
of deep understanding of the artificial neural networks, Zhao et al.
[72] proposed an overview of deep learning in palmprint recogni-
tion. Liu et al. [73] used CNN to carry out contactless recognition.
And also a novel preprocessing measure is presented [74] based on
CNN. The average accuracy of deep learning is much higher than
that of classic approaches, even reaching 100%. Thus, it is a very
promising field.
In conclusion of feature extraction, Table 3 represents the sum-
mary of the types of feature and ways to record it. As shown in
table, both fusion method and deep learning have become the pop-
ular tendency of feature extraction. Similarly, a hypothesis can be
made that the ideal way to represent feature may be the one that
fuses encoding and photo ways under the deep learning structure
in the near future.
3.2. Matching
Matching is the final step of palmprint recognition, and it is
the most essential step. The purpose of matching is to figure out
the testing palmprint image belongs to which class in the dataset.
Whether the feature is matched properly will influence the effect
of recognition system.
In this survey, how algorithms are conducted is not explained
but the matcher for comparison is mainly discussed. For differ-
ent image databases, different distances calculated will lead to dif-
ferent discrimination between the same people. Recently, many
traditional distances were still applied, such as Euclidean dis-
tance [30,122–124] , Hamming distance [26,44,47,52–54] and Chi-
square distance [32,43,46] . Some new distances were developed
as well, for example, Angular distance [115] , CW-SSIM (Complex
wavelet-structural similarity) distance [19] , Peak-to-sidelobe ratio
(PSR) [66] and Cosine Mahalanobis distance [24] are well stud-
ied. Multi-distance was also a new phenomenon. They usually
used weighted sum of multiple matchers to calculate the differ-
ence [27,35,65,115,125] . Table 4 shows the summary of distance
calculated for matching. More novel types of distance and multi-
distance newly emerged.
4. Fusion
Fusion of biometrics is the tendency with the development of
information fusion techniques. It has mutual-complement advan-
tages and can overcome drawbacks of unimodal biometrics. Several
fusion rules nowadays include minimum, maximum, sum, average,
SVM and neural networks. Fusion related to palmprint consists of
the fusion of objectives and fusion of methods for extraction and
matching.
Considering objectives used for fusion, there are many cate-
gories, such as different biometric information [4,44,108] , differ-
nt images types [21,49,128] and different features [25,31,35] . In
ddition, the level of fusion also differs in corresponding recogni-
ion system. Chief levels can be divided into four categories, pixel
4,64,95,108] , feature [21,25,52,65,128] , score [31,35,44,124,134] and
ecision level [34] . Some articles even applied two levels to
trengthen the recognition [21,49,64,108,116] . Table 5 shows the
ummary of fusion in terms of objective and level.
. Selected comparative experiments
In order to validate the performance of various methods, the
atest experimental results from the selected state-of-art works
ere presented for comparative study. It is worth noting that the
esults of them cannot be compared directly because of different
ardware condition and experimental setup.
Table 6 summarizes comparative experiments and their results.
he result indicators are RR (recognition rate) and EER. Moreover,
e choose the best experimental result in every research. Some
bbreviations used below are as follows: Anisotropic Filter (AF)
oding, LLDP (Local line directional pattern), 2D-DOST (2D discrete
rthonormal S-Transform), ST (Blocked surface type feature), LPDP
Locality preserving discriminant projections), and DCFSH (deep
onvolutional features based supervised hashing).
According to our survey, almost all the noticeable characteristics
re already considered during these years, such as shape, texture,
requency, direction, energy and phase information of a palmprint
mage. Except for innovations about the recognition or detection
tself, plenty of attention is focused on the optimization of exist-
ng strategies, thus the recognition accuracy increases like what is
emonstrated in the chart.
Essentially speaking, deep learning, subspace and encoding
ethods have superior recognition precision. Genetic features with
xcellent generalization capability can be derived by learning
ethods from large datasets, which are conveniently transplanted
nto palmprint recognition. Encoding methods did not differenti-
te which kind of feature is in the image but encode it under the
ame guideline so little feature is lost. Subspace methods train im-
ges highlighting on the features’ pattern and then obtain the pro-
ection matrix from them. Besides, deep learning approaches have
oth the biggest computational complexity and the biggest feature
ize. The reason is that too much data are needed for training so
ubstantial computing work is unavoidable. Usually, GPU is highly
equired. Similarly, structure-based methods focus only on the di-
ection and location of palm lines rather than representation of
eatures, thus the accuracy, spatial and computing complexity are
ot very ideal for extensive application. Finally, fusion approaches
olves several problems in unimodal biometrics, i.e. noisy data, il-
umination variation, partial occlusion and non-universality that
ause the system to be less accurate and secure. However, the ef-
ciency decreased meanwhile compared to other methods because
f extra work done for another fusion objective.
D. Zhong et al. / Neurocomputing 328 (2019) 16–28 25
6
6
t
d
I
i
a
d
n
a
t
D
e
e
o
o
s
c
6
fi
n
a
c
t
t
r
o
a
c
a
b
i
o
h
a
h
t
p
o
w
p
T
e
a
t
r
a
m
t
s
e
r
c
o
p
p
t
s
i
s
c
e
c
i
l
e
t
i
g
r
A
F
o
F
R
. Conclusion and discussion
.1. Conclusion of this survey
Palmprint recognition is a promising method for identity au-
hentication with the superiority of safety and stableness. In recent
ecade, it developed fast and meanwhile had many breakthroughs.
n this survey, the basic knowledge of palmprint recognition was
ntroduced firstly. Then, the developments of image acquisition
nd preprocessing were presented using a tabulation. Many new
atabases were established for simulating real world better and
ovel ways to segment ROI were also proposed. We surveyed
lmost all the valuable methods for the preprocessing stage. In
he third part, feature extraction and matching were concerned.
ifferent f eatures were obtained using three general types of
xtraction algorithms. Next, palmprint matching measures were
xplained. In the fourth part, our paper discussed the tendency
f fusion. A fact can be found that fusion improves the accuracy
f recognition and becomes increasingly popular. Finally, some
elected experiments are presented and analyzed to conduct a
omparative study as well.
.2. Suggestions on further research
After reviewing the recent works, we would like to provide our
ve suggestions and some burning research issues for further in-
ovation.
The first one should be taken into consideration is the practical
pplication. As we know, though palmprint is under research for
onsiderable years, still there are not many genuine examples
o make full use of its recognition in practice. Firstly, Rotation,
ranslation, blurring, distortion and heterogeneous data of the
equired images still resists further development [56,142] . Sec-
ndly, researchers need to figure out how to design an appropriate
lgorithm when images are captured under low or high contrast
onditions or by a contactless way. Meanwhile, image quality
ssessment is also promising to reduce the high error rate caused
y poor quality of the testing images rather than the algorithm
tself. Besides, corresponding practical datasets containing all types
f palmprint images are expected to be established, i.e. low and
igh-resolution images, 2D and 3D images and multispectral im-
ges, which will serve as a benchmark. It is better to have all the
and features included, such as vein configuration and fingerprint,
o prepare for fusion or learning method. Thirdly, due to the
rogress of Internet [12,42] , more emphasis should be placed on
nline palmprint recognition and its use in mobile devices, which
ill become a novel identification approach in online payment or
ersonal authentication.
The second promising direction is the usage of deep learning.
oo many training samples are required and it also has little gen-
ralization ability. Recently, George et al. [143] proposed a prob-
bilistic generative model called recursive cortical network (RCN)
o conduct message-passing based inference. The method unifies
ecognition, segmentation and reasoning. Excellent generalization
nd occlusion-reasoning capabilities were demonstrated. Experi-
ental results are even better than deep neural networks while
he algorithm is 300-fold more data efficient. Thus, it is a field re-
earchers may focus more on.
The third research orientation is encoding based methods. First,
ncoding way has many advantages over photo ways and plenty of
esearches have been done perfectly [54,61] . Compared to the prin-
iple of ID cards, maybe through palmprint recognition, a vector
f bitwise code called palmprint ID will be attached to everyone’s
almprint, which will be used for forensics and security protection.
The next direction is fusion. It can be used in data acquisition,
reprocessing, feature extraction and matching, leading to a bet-
er recognition performance. However, the objectives applied in fu-
ion is no more than three, maybe there should be more objectives
nvolved while total time expense shouldn’t be too high. Besides,
ometimes large amounts of information are neglected in fusion,
ausing a limited recognition rate. Finally, further researches are
xpected to consider the robustness of fusion to reduce impacts of
onstraints like illumination variation and condition changes.
The final key in palmprint recognition is liveness detection aim-
ng for high security capability. Though the palmprint cannot be
ost, forgery and duplication problems still exert huge bad influ-
nces on the recognition system. Liveness detection as a method
o detect human vital signs can prevent such attacks. Recent stud-
es like multispectral recognition [95] may be a possible solution.
Due to the limit of our perspectives, the above-mentioned sug-
estions on further research are for reference only. We welcome
eaders’ comments and suggestions.
cknowledgments
This work is supported by Grants from National Natural Science
oundation of China (No. 61105021 ), Natural Science Foundation
f Shaanxi, China (No. 2015JQ6257) and the Fundamental Research
unds for the Central Universities.
eference
[1] A .K. Jain , A . Ross , S. Prabhakar , An introduction to biometric recognition, IEEE
Trans. Circuits Syst. Video Technol. 14 (2004) 4–20 .
[2] N.F. Liu , J. Liu , Z.N. Sun , T. Tan , A code-level approach to heterogeneous irisrecognition, IEEE Trans. Inf. Forensics Secur. 12 (2017) 2373–2386 .
[3] F. Bianconi , E. Chirikhina , F. Smeraldi , C. Bontozoglou , P. Xiao , Personal iden-tification based on skin texture features from the forearm and multi-modal
imaging, Skin Res. Technol. 23 (2017) 392–398 . [4] H.G. Wang , W.Y. Yau , A. Suwandy , E. Sung , Person recognition by fusing palm-
print and palm vein images based on "Laplacianpalm" representation, Pattern
Recognit. 41 (2008) 1514–1527 . [5] D. Zhang , W. Zuo , F. Yue , A comparative study of palmprint recognition algo-
rithms, ACM Comput. Surv. 44 (2012) 1–37 . [6] A.O. Lishani , L. Boubchir , E. Khalifa , A. Bouridane , Human gait recognition
based on Haralick features, Signal Image Video Process. 11 (2017) 1123–1130 . [7] M.R. Mu , Q.Q. Ruan , S. Guo , Shift and gray scale invariant features for palm-
print identification using complex directional wavelet and local binary pat-
tern, Neurocomputing 74 (2011) 3351–3360 . [8] D. Tamrakar , P. Khanna , Noise and rotation invariant RDF descriptor for palm-
[12] O. Bingol , M. Ekinci , Stereo-based palmprint recognition in various 3D pos-tures, Expert Syst. Appl. 78 (2017) 74–88 .
[13] A. Kong , D. Zhang , M. Kamel , A survey of palmprint recognition, PatternRecognit. 42 (2009) 1408–1418 .
[14] L.K. Fei , Y. Xu , B. Zhang , X.Z. Fang , J. Wen , Low-rank representation integrated
with principal line distance for contactless palmprint recognition, Neurocom-puting 218 (2016) 264–275 .
[15] L. Leng , M. Li , C. Kim , X. Bi , Dual-source discrimination power analysis formulti-instance contactless palmprint recognition, Multimed. Tools Appl. 76
(2017) 333–354 . [16] L. Zhang , L.D. Li , A.Q. Yang , Y. Shen , M. Yang , Towards contactless palmprint
recognition: a novel device, a new benchmark, and a collaborative represen-
tation based identification approach, Pattern Recognit. 69 (2017) 199–212 . [17] J.F. Feng , C.J. Liu , H. Wang , B. Sun , High-resolution palmprint minutiae extrac-
tion based on Gabor feature, Sci. China-Inf. Sci. 57 (2014) 1–15 . [18] R.F. Wang , D. Ramos , R. Veldhuis , J. Fierrez , L. Spreeuwers , H.Y. Xu , Regional
fusion for high-resolution palmprint recognition using spectral minutiae rep-resentation, IET Biom 3 (2014) 94–100 .
[19] Z.H. Guo , D. Zhang , L. Zhang , W.H. Liu , Feature band selection for onlinemultispectral palmprint recognition, IEEE Trans. Inf. Forensic Secur. 7 (2012)
1094–1099 .
[20] B. Yang , X.Q. Xiang , D.Q. Xu , X.H. Wang , X. Yang , 3D palmprint recognitionusing shape index representation and fragile bits, Multimed. Tools Appl. 76
(2016) 15357–15375 . [21] D. Zhang , G.M. Lu , W. Li , L. Zhang , N. Luo , Palmprint recognition using 3-D in-
formation, IEEE Trans. Syst. Man Cybern. Part C-Appl. Rev. 39 (2009) 505–519 .
26 D. Zhong et al. / Neurocomputing 328 (2019) 16–28
[22] J.R. Cui , 2D and 3D palmprint fusion and recognition using PCA plus TPTSRmethod, Neural Comput. Appl. 24 (2014) 497–502 .
[23] X.Y. Jing , S. Li , W.Q. Li , Y.F. Yao , C. Lan , J.S. Lu , J.Y. Yang , Palmprint and facemulti-modal biometric recognition based on SDA-GSVD and its kernelization,
Sensors 12 (2012) 5551–5571 . [24] M. Chaa , N.E. Boukezzoula , A. Attia , Score-level fusion of two-dimensional
and three-dimensional palmprint for personal recognition systems, J. Elec-tron. Imaging 26 (2017) 013018 .
[25] J.F. Dai , J. Zhou , Multifeature-based high-resolution palmprint recognition,
IEEE Trans. Pattern Anal. Mach. Intell. 33 (2011) 945–957 . [26] Z.H. Guo , D. Zhang , L. Zhang , W.M. Zuo , G.M. Lu , Empirical study of light
source selection for palmprint recognition, Pattern Recognit. Lett. 32 (2011)120–126 .
[27] W. Jia , R.X. Hu , J. Gui , Y. Zhao , X.M. Ren , Palmprint recognition across differentdevices, Sensors 12 (2012) 7938–7964 .
[28] M.R. Mu , Q.Q. Ruan , Mean and standard deviation as features for palmprint
recognition based on Gabor filters, Int. J. Pattern Recognit. Artif. Intell. 25(2011) 491–512 .
[33] M. Aykut , M. Ekinci , AAM-based palm segmentation in unrestricted back-grounds and various postures for palmprint recognition, Pattern Recognit.
Lett. 34 (2013) 955–962 .
[34] J.Y. Ni , J. Luo , W.B. Liu , 3D palmprint recognition using Dempster-Shafer fu-sion theory, J. Sens. (2015) 252086 .
[35] B. Yang , X.H. Wang , J.L. Yao , X. Yang , W.H. Zhu , Efficient local representationsfor three-dimensional palmprint recognition, J. Electron. Imaging 22 (2013)
043040 . [36] W. Li , B. Zhang , L. Zhang , J.Q. Yan , Principal line-based alignment refinement
for palmprint recognition, IEEE Trans. Syst. Man Cybern. Part C-Appl. Rev. 42
(2012) 1491–1499 . [37] D. Zhang , V. Kanhangad , N. Luo , A. Kumar , Robust palmprint verification using
2D and 3D features, Pattern Recognit. 43 (2010) 358–368 . [38] X. Pan , Q.Q. Ruan , Palmprint recognition using Gabor-based local invariant
features, Neurocomputing 72 (2009) 2040–2045 . [39] L.L. Shen , S.P. Wu , S.H. Zheng , Z. Ji , Embedded palmprint recognition system
using OMAP 3530, Sensors 12 (2012) 1482–1493 .
[40] H. Ota , S. Aoyama , R. Watanabe , K. Ito , Y. Miyake , T. Aoki , Implementationand evaluation of a remote authentication system using touchless palmprint
recognition, Multimedia Syst 19 (2013) 117–129 . [41] Z.N. Sun , L.B. Wang , T.N. Tan , Ordinal feature selection for iris and palmprint
recognition, IEEE Trans. Image Process. 23 (2014) 3922–3934 . [42] J.S. Kim , G. Li , B. Son , J. Kim , An empirical study of palmprint recognition for
mobile phones, IEEE Trans. Consum. Electron. 61 (2015) 311–319 . [43] D. Tamrakar , P. Khanna , Blur and occlusion invariant palmprint recognition
with block-wise local phase quantization histogram, J. Electron. Imaging 24
(2015) 043006 . [44] A. Meraoumia , S. Chitroub , A. Bouridane , Palmprint and finger-knuckle-print
for efficient person recognition based on log-Gabor filter response, Analog In-tegr. Circuits Process. 69 (2011) 17–27 .
[45] F. Yue , W.M. Zuo , D. Zhang , K.Q. Wang , Orientation selection using modifiedFCM for competitive code-based palmprint recognition, Pattern Recognit. 42
(2009) 2841–2849 .
[46] M. Hammami , S. Ben Jemaa , H. Ben-Abdallah , Selection of discriminativesub-regions for palmprint recognition, Multimed. Tools Appl. 68 (2014)
1023–1050 . [47] G.S. Badrinath , P. Gupta , Palmprint based recognition system using phase-dif-
ference information, Futur. Gener. Comp. Syst. 28 (2012) 287–305 . [48] K. Tiwari , D.K. Arya , G.S. Badrinath , P. Gupta , Designing palmprint based
recognition system using local structure tensor and force field transformation
for human identification, Neurocomputing 116 (2013) 222–230 . [49] R. Raghavendra , C. Busch , Novel image fusion scheme based on dependency
[50] D.F. Hong , J. Su , Q.G. Hong , Z.K. Pan , G.D. Wang , Blurred palmprint recognitionbased on stable-feature extraction using a Vese-Osher decomposition model,
PLoS One 9 (2014) e101866 .
[51] D. Tamrakar , P. Khanna , Kernel discriminant analysis of block-wise Gaussianderivative phase pattern histogram for palmprint recognition, J. Vis. Commun.
Image Represent. 40 (2016) 432–448 . [52] D.F. Hong , W.Q. Liu , J. Su , Z.K. Pan , G.D. Wang , A novel hierarchical approach
for multispectral palmprint recognition, Neurocomputing 151 (2015) 511–521 .
[53] L. Zhang , H.Y. Li , J.Y. Niu , Fragile bits in palmprint recognition, IEEE Signal
Process. Lett. 19 (2012) 663–666 . [54] H.P. Chen , An efficient palmprint recognition method based on block dominat
orientation code, Optik 126 (2015) 2869–2875 . [55] L.K. Fei , B. Zhang , Y. Xu , L.P. Yan , Palmprint recognition using neighboring di-
rection indicator, IEEE T. Hum.-Mach. Syst. 46 (2016) 787–798 .
[56] D.F. Hong , W.Q. Liu , X. Wu , Z.K. Pan , J. Su , Robust palmprint recognitionbased on the fast variation Vese-Osher model, Neurocomputing 174 (2016)
999–1012 . [57] W. Jia , R.X. Hu , Y.K. Lei , Y. Zhao , J. Gui , Histogram of oriented lines for palm-
print recognition, IEEE Trans. Syst. Man Cybern. -Syst. 44 (2014) 385–395 . [58] L.M. Cui , Y.T. Wei , Y.Y. Tang , H. Li , Gabor-based tensor local discriminant em-
bedding and its application on palmprint recognition, Int. J. Wavelets Mul-tiresolut. Inf. Process. 8 (2010) 327–342 .
[59] M. Laadjel , S. Al-Maadeed , A. Bouridane , Combining fisher locality preserving
projections and passband DCT for efficient palmprint recognition, Neurocom-puting 152 (2015) 179–189 .
[60] A .A . Altun , A combination of genetic algorithm, particle swarm optimiza-tion and neural network for palmprint recognition, Neural Comput. Appl. 22
(2013) S27–S33 . [61] L.K. Fei , Y. Xu , W.L. Tang , D. Zhang , Double-orientation code and nonlin-
ear matching scheme for palmprint recognition, Pattern Recognit 49 (2016)
89–101 . [62] M.R. Mu , Q.Q. Ruan , Region covariance matrices as feature descriptors for
palmprint recognition using Gabor features, Int. J. Pattern Recognit. Artif. In-tell. 25 (2011) 513–528 .
[63] Q.E. Wu , Z.W. Chen , R.J. Han , C.X. Yang , Y.H. Du , Y.X. Zheng , W.F. Cheng , Apalmprint recognition approach based on image segmentation of region of
[64] A. Bouchemha , N. Doghmane , M.C. Nait-Hamoud , A. Nait-Ali , Multispectralpalmprint recognition methodology based on multiscale representation, J.
Electron. Imaging 24 (2015) 043005 . [65] M.D. Bounneche , L. Boubchir , A. Bouridane , B. Nekhoul , A. Ali-Cherif , Multi-
-spectral palmprint recognition based on oriented multiscale log-Gabor fil-ters, Neurocomputing 205 (2016) 274–286 .
[66] W. Jia , B. Zhang , J.T. Lu , Y.H. Zhu , Y. Zhao , W.M. Zuo , H.B. Ling , Palmprint
recognition based on complete direction representation, IEEE Trans. ImageProcess. 26 (2017) 4 483–4 498 .
[67] M. Tabejamaat , A. Mousavi , A coding-guided holistic-based palmprint recog-nition approach, Multimed. Tools Appl. 76 (2017) 7731–7747 .
[68] L. Zhang , H.Y. Li , Encoding local image patterns using Riesz transforms: withapplications to palmprint and finger-knuckle-print recognition, Image Vis.
Comput. 30 (2012) 1043–1051 .
[69] Y. Yan , Y.J. Zhang , Discriminant projection embedding for face and palmprintrecognition, Neurocomputing 71 (2008) 3534–3543 .
[70] W.M. Zuo , H.Z. Zhang , D. Zhang , K.Q. Wang , Post-processed LDA for faceand palmprint recognition: what is the rationale, Signal Process. 90 (2010)
2344–2352 . [71] S. Minaee , W. Yao , Palmprint recognition using deep scattering convolutional
network, in: Proceedings of the 2017 IEEE International Symposium on Cir-
cuits and Systems (ISCAS), 2017, pp. 1–4 . [72] Z. Dandan , X. Pan , X. Luo , X. Gao , Palmprint recognition based on deep learn-
ing, in: Proceedings of the 6th International Conference on Wireless, Mobileand Multi-Media (ICWMMN 2015), 2015, pp. 214–216 .
[73] D. Liu , D. Sun , Contactless palmprint recognition based on convolutional neu-ral network, in: Proceedings of 2016 IEEE 13th International Conference on
Signal Processing, 2016, pp. 1363–1367 . [74] X. Bao , Z. Guo , Extracting region of interest for palmprint by convolutional
neural networks, in: Proceedings of the 2016 Sixth International Conference
on Image Processing Theory, Tools and Applications, 2016, pp. 1–6 . [75] X. Wang , L. Lei , M. Wang , Palmprint verification based on 2D-Gabor wavelet
and pulse-coupled neural network, Knowl.-Based Syst. 27 (2012) 451–455 . [76] W. Yang , L. Wang , Research of palmprint identification method using Zernike
moment and neural network, in: Proceedings of the 2010 Sixth InternationalConference on Natural Computation, 2010, pp. 1310–1313 .
[77] L. Huang , N. Li , in: Palmprint Recognition Using Polynomial Neural Net-
[78] X. Cheng , S. Wang , A new palmprint identification technique based on atwo: stage neural network classifier, in: Proceedings of the 2008 International
Conference on Computer Science and Software Engineering, 2008, pp. 957–960 .
[79] P. Vu , D. Xu , IEEE, palmprint recognition based on modified DCT features and
RBF neural network, in: Proceedings of the 2008 International Conference onMachine Learning and Cybernetics, 2008, pp. 2982–2986 .
[80] J. Lu , Y. Zhao , Y. Xue , J. Hu , Palmprint recognition via locality preserving pro-jections and extreme learning machine neural network, in: Proceedings of the
2008 9th International Conference on Signal Processing, 2008, pp. 2097–2100 .[81] Z. Chen , L. Huang , Palmprint recognition based on ICA and BP neural network,
J. Beijing Univ. Aeronaut. Astronaut. 34 (2008) 290–294 .
[82] D. Zhang , W.M. Zuo , F. Yue , A comparative study of palmprint recognitionalgorithms, ACM Comput. Surv. 44 (2012) 37 .
[83] X.R. Peng , Y.M. Tian , J.Q. Wang , IEEE, A survey of palmprint feature extractionalgorithms, in: Proceedings of the 2013 Fourth International Conference on
Intelligent Systems Design and Engineering Applications, 2013, pp. 57–63 . [84] M.K. Kim , Palmprint recognition based on line and slope orientation features,
J. Inf. Sci. Eng. 27 (2011) 1219–1232 .
[85] M. Tabejamaat , A. Mousavi , Concavity-orientation coding for palmprint recog-nition, Multimed. Tools Appl. 76 (2017) 9387–9403 .
[86] W. Huang , X. Lin , X. Dai , A novel approach for palmprint ridges features ex-traction, in: Proceedings of the 2009 2nd International Congress on Image
D. Zhong et al. / Neurocomputing 328 (2019) 16–28 27
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[87] J. Dai , J. Feng , J. Zhou , Robust and efficient ridge-based palmprint matching,IEEE Trans. Pattern Anal. Mach. Intell. 34 (2012) 1618–1632 .
[88] J. Chen , Z. Guo , Palmprint matching by minutiae and ridge distance, cloudcomputing and security, ICCCS 2016 (2016) 371–382 .
[89] C. Li , F. Liu , Y. Zhang , A principal palm-line extraction method for palmprintimages based on diversity and contrast, in: Proceedings of the 3rd Interna-
tional Congress on Image and Signal Processing, 2010, pp. 1772–1777 . [90] A .S. Parihar , A . Kumar , O.P. Verma , A . Gupta , P. Mukherjee , D. Vatsa , IEEE,
point based features for contact-less palmprint images, in: Proceedings of the
IEEE International Conference on Technologies for Homeland Security, 2013,pp. 165–170 .
[91] Y.T. Luo , L.Y. Zhao , B. Zhang , W. Jia , F. Xue , J.T. Lu , Y.H. Zhu , B.Q. Xu , Localline directional pattern for palmprint recognition, Pattern Recognit 50 (2016)
26–44 . [92] A. Bruno , P. Carminetti , V. Gentile , M. La Cascia , E. Mancino , IEEE, palm-
print principal lines extraction, in: Proceedings of the 2014 IEEE Workshop
on Biometric Measurements and Systems for Security and Medical Applica-tions, 2014, pp. 50–56 .
[93] M.M.H. Ali , P. Yannawar , A.T. Gaikwad , IEEE, study of edge detection meth-ods based on palmprint lines, in: Proceedings of the 2016 International
Conference on Electrical, Electronics, and Optimization Techniques, 2016,pp. 1344–1350 .
[94] X.B. Xu , D.Y. Zhang , X.M. Zhang , Y.J. Cao , Palmprint recognition based on
discrete curvelet transform and support vector machine, J. Infrared Millim.Waves 28 (2009) 456–460 .
[95] X.B. Xu , L.B. Lu , X.M. Zhang , H.M. Lu , W.Y. Deng , Multispectral palmprintrecognition using multiclass projection extreme learning machine and digi-
tal shearlet transform, Neural Comput. Appl. 27 (2016) 143–153 . [96] R. Gayathri , P. Ramamoorthy , Automatic palmprint identification based on
high order Zernike moment, Am. J. Appl. Sci. 9 (2012) 759–765 .
[97] Y. Ma , J. Sun , Palmprint recognition based on 2DPCA-moment invariant, in:Proceedings of the Fifth International Conference on Image and Graphics,
2009, pp. 149–155 . [98] L.C. Deepika , A. Kandaswamy , C. Vimal , B. Satish , Palmprint authentication
using modified Legendre moments, in: Proceedings of the International Con-ference and Exhibition on Biometrics Technology, 2010, pp. 164–172 .
[99] Y. Li , G. Wu , J. Li , G. Dai , Palmprint matching algorithm based on atrous–
Contourlet transform and invariant moments, J. Comput. Appl. 30 (2010)956–959 .
100] I.S. Hussein , M.J. Nordin , Palmprint identification using invariant moments al-gorithm based on wavelet transform, Adv. Comput. Commun. Eng. Technol.
315 (2015) 905–914 . [101] X. Bai , N. Gao , Z. Zhang , D. Zhang , 3D palmprint identification combining
blocked ST and PCA, Pattern Recognit. Lett. 100 (2017) 89–95 .
[102] J. Li , Research on applied technology of palmprint recognition based on Fisherlinear discriminant and improved PCA algorithm, Adv. Res. Mater. Sci., Envi-
ron. Sci. Comput. Sci. III 886 (2014) 515–518 . [103] K. Li , Y. Zhang , H. Liu , S. Sun , Palmprint recognition based on modular
PCA and LS-SVM, in: Proceedings of the 2015 4th International Confer-ence on Computer, Mechatronics, Control and Electronic Engineering, 2015,
pp. 566–571 . 104] J. Cui , Multispectral palmprint recognition using image-based linear discrim-
inant analysis, Int. J. Biom. 4 (2012) 106–115 .
[105] S. Xu , J. Suo , J. Ding , Improved linear discriminant analysis based on two-di-mensional Gabor for palmprint recognition, in: Proceedings of the 2011 In-
ternational Conference of Soft Computing and Pattern Recognition, 2011,pp. 157–160 .
106] S. Xu , J. Ding , Palmprint image processing and linear discriminant analysismethod, J. Multimedia 7 (2012) 269–276 .
[107] L. Shang , P. Su , G. Dai , Y. Gu , Z. Zhao , Palmprint recognition method us-
ing WTA-ICA based on 2DPCA, Adv. Intell. Comput. Theories Appl. (2010)250–257 .
[108] J.W. Lu , Y.P. Tan , Improved discriminant locality preserving projections forface and palmprint recognition, Neurocomputing 74 (2011) 3760–3767 .
[109] J. Lu , Y. Zhao , Y. Xue , J. Hu , Palmprint recognition via locality preservingprojections and extreme learning machine neural network, in: Proceedings
of the 9th International Conference on Signal Processing, 2008, pp. 2096–
2099 . [110] J. Guo , L. Gu , Y. Liu , Y. Li , J. Zeng , Palmprint recognition based on kernel local-
ity preserving projections, in: Proceedings of the 3rd International Congresson Image and Signal Processing, 2010, pp. 1909–1913 .
[111] S. Li , Palmprint recognition method based on a new kernel sparse representa-tion method, intelligent computing theories and technology, in: Proceedings
of the 9th International Conference, ICIC 2013, 2013, pp. 515–523 .
[112] Y. Zhang , M. Qi , L. Shang , Palmprint recognition based on two-dimensionalGabor wavelet transform and two-dimensional principal component analysis,
Adv. Intell Comput. (2011) 405–411 . [113] A. Younesi , M.C. Amirani , IEEE, palmprint identification via GLCM of con-
tourlet transform, in: Proceedings of the 2013 First International Conferenceon Communications Signal Processing, and Their Applications, 2013, pp. 1–5 .
[114] S. Saedi , N.M. Charkari , Palmprint authentication based on discrete orthonor-
mal S-Transform, Appl. Soft Comput. 21 (2014) 341–351 . [115] W.M. Zuo , F. Yue , D. Zhang , On accurate orientation extraction and appropri-
ate distance measure for low-resolution palmprint recognition, Pattern Recog-nit. 44 (2011) 964–972 .
[116] A. Meraoumia , S. Chitroub , A. Bouridane , 2D and 3D palmprint information,PCA and HMM for an improved person recognition performance, Integr. Com-
put.-Aided Eng. 20 (2013) 303–319 . [117] R. Raghavendra , C. Busch , Texture based features for robust palmprint recog-
nition: a comparative study, EURASIP J. Inf. Secur. (2015) 1–9 . [118] A. Saliha , D.H. Nabil , B. Karima , K. Mouloud , B. Ahmed , Palmprint recogni-
tion system: case study with different databases, J. Inf. Assur. Secur. 10 (2015)130–138 .
[119] R. Mokni , M. Kherallah , Palmprint recognition through the fractal dimension
estimation for texture analysis, Int. J. Biom. 8 (2016) 254–274 . 120] X.M. Guo, W.D. Zhou, A Novel Dictionary Learning model with PT-HLBP for
for palmprint recognition, Opt. Eng. 51 (2012) 1–9 . 124] X. Jia , J. Cui , D. Xue , F. Pan , Near Infrared vein image acquisition system
based on image quality assessment, in: Proceedings of the 2011 Interna-tional Conference on Electronics, Communications and Control (ICECC), 2011,
pp. 922–925 .
125] Z.H. Liu , J.X. Pu , T. Huang , Y. Qiu , A novel classification method for palmprintrecognition based on reconstruction error and normalized distance, Appl. In-
tell. 39 (2013) 307–314 . 126] J. Gui , W. Jia , L. Zhu , S.L. Wang , D.S. Huang , Locality preserving discriminant
projections for face and palmprint recognition, Neurocomputing 73 (2010)2696–2707 .
[127] J.J. Wen , Y. Chen , J.X. Mi , A palmprint recognition method based on multi-step
representation, Optik 124 (2013) 5727–5731 . 128] Y. Xu , Z.Z. Fan , M.N. Qiu , D. Zhang , J.Y. Yang , A sparse representation method
of bimodal biometrics and palmprint recognition experiments, Neurocomput-ing 103 (2013) 164–171 .
129] S.W. Zhang , X.X. Gu , Palmprint recognition based on the representation in thefeature space, Optik 124 (2013) 5434–5439 .
130] J.J. Wen , J.R. Cui , Z.H. Lai , J.X. Mi , X.T. Guo , Optimal representation set con-
struction with a competitive scheme for palmprint and face recognition, Int.J. Comput. Math. 91 (2014) 2341–2358 .
[131] X.B. Xu , X.M. Zhang , L.B. Lu , W.Y. Deng , K.L. Zuo , Fast near-infrared palmprintrecognition using nonnegative matrix factorization extreme learning machine,
Opt. Appl. 44 (2014) 285–298 . 132] J.R. Cui , J.J. Wen , Z.Z. Fan , Appearance-based bidirectional representation for
[133] B. Yang , X.Q. Xiang , D.Q. Xu , X.H. Wang , X. Yang , 3D palmprint recognitionusing shape index representation and fragile bits, Multimed. Tools Appl. 76
(2017) 15357–15375 . 134] J.R. Cui , Multispectral fusion for palmprint recognition, Optik 124 (2013)
3067–3071 . [135] E. Perumal , S. Ramachandran , A multimodal biometric system based on palm-
print and finger knuckle print recognition methods, Int. Arab J. Inf. Technol.12 (2015) 118–128 .
136] L.B. Lu , X.M. Zhang , X.B. Xu , D.P. Shang , Multispectral image fusion for illumi-
nation-invariant paimprint recognition, PLoS One 12 (2017) 1–22 . [137] G.S. Badrinath , N.K. Kachhi , P. Gupta , Verification system robust to occlu-
sion using low-order Zernike moments of palmprint sub-images, Telecom-mun. Syst. 47 (2011) 275–290 .
138] X.F. Bai , N. Gao , Z.H. Zhang , D. Zhang , 3D palmprint identification combiningblocked ST and PCA, Pattern Recognit. Lett. 100 (2017) 89–95 .
139] J. Gui , W. Jia , L. Zhu , S.-L. Wang , D.-S. Huang , Locality preserving discriminant
projections for face and palmprint recognition, Neurocomputing 73 (2010)2696–2707 .
[140] M. Farmanbar , O. Toygar , Feature selection for the fusion of face and palm-print biometrics, Signal Image Video Process. 10 (2016) 951–958 .
[141] J. Cheng , Q. Sun , J. Zhang , Q. Zhang , Supervised Hashing with Deep Convolu-tional Features for Palmprint Recognition, Springer International Publishing,
Cham, 2017, pp. 259–268 .
[142] G. Li , J. Kim , Palmprint recognition with local micro-structure tetra pattern,Pattern Recognit 61 (2017) 29–46 .
[143] D. George, W. Lehrach, K. Kansky, M. Lázaro-Gredilla, C. Laan, B. Marthi,X. Lou, Z. Meng, Y. Liu, H. Wang, A. Lavin, D.S. Phoenix, A generative vision
model that trains with high data efficiency and breaks text-based CAPTCHAs,Science 358 (6368) (2017) 1–9 . Article ID eaag2612, doi: 10.1126/science.
aag2612 .
Dexing Zhong received his Bs.Sc. and Ph.D. degrees fromXi’an Jiaotong University in 2005 and 2010, respectively.
He is an associate professor in School of Electronic andInformation Engineering, Xi’an Jiaotong University, China.
He was a visiting scholar with University of Illinois atUrbana-Champaign, United States. His main research in-