Palm vein pattern-based biometric recognition system

Abstract

Palm vein-based biometric authentication system aims to recognise individuals from their unique palm vein structure which is next to impossible to duplicate owing to the fact that palm veins are present in the subsurface of the skin and not apparent under visual light. The aim of the proposed work is to develop a low cost but efficient system for acquiring images of the veins, processing these images and matching using various algorithms. Images have been acquired using a web camera and an infrared LED illumination that highlights the veins. The region of interest (ROI) is extracted from the images and then processed. Three techniques for matching are proposed. Principal component analysis (PCA), 2D-wavelet based feature and template designed exclusively for palm vein ROI is applied over ROI for matching. The accuracy of the each algorithm is deduced to compare the three algorithms. The highest accuracy achieved is 93.54% using template matching technique.

Full text

Int. J. Computer Applications in Technology, Vol. 51, No. 2, 2015 105
Copyright © 2015 Inderscience Enterprises Ltd.
Palm vein pattern-based biometric recognition
system
Gunjan Shah, Sagar Shirke, Sonam Sawant
and Yogesh H. Dandawate*
Department of Electronics and Telecommunications,
Vishwakarma Institute of Information Technology,
Pune, 411048, India
Email: shah.gunjan27@yahoo.com
Email: sagar_shirke12@yahoo.com
Email: sonamsawant@yahoo.co.in
Email: yhdandawate@gmail.com
Email: yogesh.dandawate@viit.ac.in
*Corresponding author
Abstract: Palm vein-based biometric authentication system aims to recognise individuals from
their unique palm vein structure which is next to impossible to duplicate owing to the fact that
palm veins are present in the subsurface of the skin and not apparent under visual light. The aim
of the proposed work is to develop a low cost but efficient system for acquiring images of the
veins, processing these images and matching using various algorithms. Images have been
acquired using a web camera and an infrared LED illumination that highlights the veins. The
region of interest (ROI) is extracted from the images and then processed. Three techniques for
matching are proposed. Principal component analysis (PCA), 2D-wavelet based feature and
template designed exclusively for palm vein ROI is applied over ROI for matching. The
accuracy of the each algorithm is deduced to compare the three algorithms. The highest accuracy
achieved is 93.54% using template matching technique.
Keywords: biometrics; palm vein patterns; infrared LED; 2D wavelet transform; template
matching; PCA; principal component analysis.
Reference to this paper should be made as follows: Shah, G., Shirke, S., Sawant, S.
and Dandawate, Y.H. (2015) ‘Palm vein pattern-based biometric recognition system’,
Int. J. Computer Applications in Technology, Vol. 51, No. 2, pp.105–111.
Biographical notes: Gunjan Shah is an Undergraduate student in Electronics and
Telecommunication Engineering Department at Vishwakarma Institute of Information
Technology, Pune. She recently received Bachelor of Engineering in Electronics and
Telecommunication Engineering form University of Pune.
Sagar Shirke is an Undergraduate student in Electronics and Telecommunication Engineering
Department at Vishwakarma Institute of Information Technology, Pune. He recently received
Bachelor of Engineering in Electronics and Telecommunication Engineering form University of
Pune.
Sonam Sawant is an Undergraduate student in Electronics and Telecommunication Engineering
Department at Vishwakarma Institute of Information Technology, Pune. She recently received
Bachelor of Engineering in Electronics and Telecommunication Engineering form University of
Pune.
Yogesh H. Dandawate received his Bachelor of Engineering from University of Pune (India) in
1991, Masters of Engineering from Gulbarga University (India) in 1998 and PhD in Electronics
and Telecommunications Engineering in 2009. He started his carrier as Lecturer in Engineering
in 1992. Presently, he is working as Professor in Electronics and Telecommunications
Engineering Department at Vishwakarma Institute of Information Technology, Pune. He has
22 years of teaching experience and has published more than 37 papers in reputed national
and International Conferences/referred Journals. His areas of interests are signal and image
processing, embedded systems and soft computing. He is a reviewer and editorial board member
of reputed journals in India and abroad. He is also working on several research projects. He is
senior member of IEEE, and Fellow member of IETE, India.

106 G. Shah et al.
1 Introduction
Biometrics is a method of recognising individuals based
on their physiological or behavioural characteristic.
The advantage of biometric authentication is that the
biometric data are based on physical characteristics that stay
constant throughout one’s lifetime and are difficult (some
more than others) to duplicate or change. Duplication
(or copying) of physical characteristics is difficult when
compared with older techniques of keys, smart cards, and
personal identification numbers (PINs) or passwords. Thus,
biometrics has been used as identification technique for
humans for several years.
The biometric systems which we come across
frequently nowadays are fingerprints, voice, hand geometry,
palm print, iris or pupil, palm vein, finger vein and many
more. Fingerprint is an easy technique which we come
across more often than the others even at public places not
just for security purpose but also for attendance or as a
substitute for signatures. Although the other systems have
also been used quite regularly for security purpose and none
of these are full proof and insusceptible to errors. All these
biometric techniques have their pros and cons.
(Bhattacharyya et al., 2009; Zhang, 2000). Fingerprints as
we know can be obtained from any surface, thus duplicated
using even cello tapes thus spoofing is quite an easy task.
An individual who is colour-blind cannot pass the test of iris
recognition and this can be duplicated by simply using
lenses. Measurement accuracy of retinal scan can be
affected by a disease such as cataracts. Voice recognition
suffers from disadvantages related to storage space,
i.e., memory usage, recording difficulties and most
importantly noise. Hand geometry is not known to be
unique and can be duplicated easily, owing to this it is not
used in identification systems regularly and on a large scale.
Personal verification has become an important and high
demand technique for security access systems over the last
decade. Shape of the subcutaneous vascular tree of the
dorsal hand contains information that is capable of
authenticating the individual to a reasonable accuracy for
automatic personal authentication purpose. Wang et al.
(2007) suggested the vein pattern biometric with infrared
imaging as a potential biometric. Vein pattern proves to be
advantageous compared with the other biometric techniques
because it cannot be forged easily as it lies in the
subcutaneous layer. Sarkar et al. (2010) have mentioned
other advantages of palm veins like accuracy and reliability,
contact-less, cost-effective and usability.
For more accurate recognition of the patterns images
obtained from the sensors must be of good quality.
Many techniques have been used for the capture of the
palm vein images and matching of the pattern for the
purpose of identification and recognition. In this paper, we
are proposing a simple and low-cost technique for an
effective palm vein image capture, computationally efficient
region of interest (ROI) technique for better accuracy and
different techniques such as template matching, Wavelet
transform and principle component analysis (PCA) for
recognition/identification of an individual. Use 2D discrete
wavelet transform (DWT) is attempted. The performance of
these techniques is compared. It has been found that PCA
and template matching techniques provide better accuracy
than wavelets. Accuracy in template matching is owing to
better imaging and ROI extraction in comparison with the
different pattern matching techniques proposed so far.
The rest of the paper is organised as follows. Section 2
present a principle of the palm vein image capture
and techniques used along with proposed one for dorsal
palm vein capture. The ROI extraction techniques
along with proposed technique, the template preparation,
feature/pattern extraction using PCA and DWT are
discussed in Section 3. Results of experimentation using our
own generated database are presented in Section 4 and
finally Section 5 concludes the paper.
2 Image capture system
The principle used for capturing dorsal palm veins and the
proposed setup is discussed in following sections.
2.1 Principle of palm vein image capture
The palm veins are not visible clearly in normal light. The
visibility of veins depends on factors like age, subcutaneous
fat, humidity and temperature. Through our experiments we
also presume that the visibility of veins in infrared light
depends on the haemoglobin levels in the blood. The
haemoglobin contained in the blood absorbs the infrared
light for a larger amount of time than the other parts
of the hand and thus using an infrared camera we can
capture the images in which the veins are highlighted. For
the purpose of the use of proper light source for imaging the
penetration of different wavelengths into the skin is studied.
Figure 1 shows the depth of penetration at different
wavelengths.
Figure 1 Penetration of different wavelengths into skin
Source: Bashkatov et al. (2005)
There are two types of infrared imaging techniques namely,
near infrared (NIR) and far infrared (FIR). NIR is less
sensitive to temperature and humidity hence proves to be a
better source of illumination. Wang and Leedham (2006)
have presented the advantages and disadvantages of both the
techniques. NIR has been used in our system.

Palm vein pattern-based biometric recognition system 107
There are two methods for image capturing namely,
reflection and penetration. When reflection is used the
infrared illumination source and camera are on the same
side of the hand (generally LEDs are encircling the camera),
whereas for penetration the hand is between the illumination
source and the camera. Most of the work on this biometric
modality has preferred the technique of reflection. Han and
Lee (2012) have also employed reflection for capturing
images. 2D Gabor filter is used to obtain vein pattern
at low resolution. Low resolution is used to prevent noise.
Jia et al. (2012) have used two LED array lamps for
illumination and they have captured the images using
reflection. Further Radon transform is used to get positional
and directional pattern of the veins. After experimentation
we observed that we obtained better images of dorsal veins
when we used penetration. From the literature cited above
infrared or NIR that is more suitable and harmless to skin is
required. Either a special IR-based camera can be used that
is costly or some modifications are required are required
when a normal digital camera in visible range is used.
2.2 Palm vein image capture setup
Designing a good-quality image capture system is of utmost
importance. Obtaining good contrast, low noise and clear
image reduces the burden on pre-processing and thus helps
us to concentrate on the matching techniques.
Lee (2012) has mentioned the use of DSP camera with
LEDs arranged circularly around it. Optical infrared filter
(Hoya RM80 IR filter) has also been used to enhance the
quality of the images. Optical infrared filter was placed
between the hand and the camera lens.
Wang et al. (2007) used Hitachi KP-F2A infrared CCD
camera, NIR and FIR LEDs with Hoya RM80 IR filter.
Jia et al. (2012) used two LED lamps inclined at an angle on
opposite ends and a CMOS sensor camera for image
capturing. Nandini et al. (2012) suggested use of a simple
camera under normal conditions of temperature and
lightening. On normal conditions grey-scale discrimination
of vein image is very small. The pre-processing
requirements are operations such as segmentation, filtering,
thinning and Hough transformation.
Bu et al. (2012) have recommended the use of CD
cameras. The NIR LEDs and camera are installed on top
and bottom of the hand to capture both dorsal and palm
veins. Sarkar et al. (2010) have used an IR cold source as a
solid state array of 24 LEDs. Monochrome CCD fitted
camera is used with IR filter to image the back of the hand.
The LEDs used are of 750 nm wavelength. Kumar and
Prathyusha (2008) used low-cost NIR camera, traditionally
used for surveillance is employed for contactless image
acquisition. The NIR (850 nm) are circularly distributed
around the camera. Even though wavelengths more than 850
nm penetrates at more depth, it may harm the skin.
Sometimes ultra violet LEDs can also be used. On the basis
of the studies, we have used 850 nm LEDs for our capture
system. The aim was to develop a low cost but still
providing better quality imaging system. For the fulfilment
we have used a regular webcam that is readily available
in the market and modified it to make it IR sensitive.
The iBall face to face c12.0 (1.3 Mega Pixel) webcam
is modified to sense only (or mostly) the infrared light.
For this purpose, infrared filter has been removed from
the webcam. Figure 2 shows steps for removal of IR filter
from the web cam. The scale beside camera indicates size of
the IR filter.
Figure 2 Steps for removing IR filter from the web cam
(see online version for colours)
For illumination 44 high powered NIR LEDs, SFH4550
by Osram have been arranged in a matrix pattern to
form an illumination source. LEDs have to be arranged
in a way to compensate for the non-uniform thickness
of the palm. Arrangement for light intensity control
has been provided. Acrylic and etched glass have been
used for uniform illumination. We also tried to place
camera film to avoid entering of excess light entering
from sides of palm into the camera. Figure 3 shows the
arrangement of LEDs and the setup for capturing of palm
vein images.
Figure 3 LED arrangement and palm vein image capture setup
(see online version for colours)
2.2.1 Database
The database consists of 62 images. The captured images
are 24 bit RGB with a resolution of 640 × 480 pixels. The
distance between the hand and the camera is approximately
kept 3 inches. The position of the hand is restricted while
capturing images. Three images of dorsal veins from left
hand are captured for each individual.

108 G. Shah et al.
3 Pre-processing, feature extraction and
recognition
The first step towards pre-processing is extraction of ROI.
It is required because area of same size is required
for matching templates or any statistical parameters. Proper
extraction of ROI is first step towards effective recognition.
The extracted ROI is enhanced using various image
enhancement algorithms. The features are extracted using
different techniques and matched by using different distance
measures.
3.1 ROI extraction and enhancement
Zhou and Kumar (2011) have suggested the extraction of
ROI from the hand image using webs. Then the ROIs are
processed using histogram equalisation and adaptive
histogram equalisation and resized to the required size using
bicubic interpolation. Neighbourhood Radon Transform has
been used for matching. ROI can also be extracted using
valley points between ring finger and small finger (Lee,
2012; Wang et al., 2007). Local thresholding and
skeletonisation is then performed. Jia et al. (2012) extracted
the ROI by selecting knuckle points. Because there is too
much noise in infrared vein image and the image contrast is
very low, median filter and contrast enhancement algorithm
are applied to the vein image, and normalised. Ghosh and
Dutta (2102) have used three processes namely,
• vascular pattern pointer algorithm (VPPA)
• vascular pattern G-B conversion algorithm (VPGBCA)
• vascular pattern thinner algorithm (VPTA).
During first process a near-infrared image is converted into
a grey-scale image. In the second process, the grey-scale
image is again converted into a binary image. Lastly, the
processed binary image is finely thinned to get a proper
thinned image. This VPTA process gives an edge to enrich
the level of security of perfect biometric authentication to
maximum level.
Nandini et al. (2012) recommend a dynamic threshold-
based segmentation process is carried out which subdivides
the image into its constituent regions. The vein patterns are
extracted according to the threshold selected. To enhance
the quality of vein patterns obtained, different filters was
applied on these segmented vein patterns like Wiener filter
which help in preserving edges and other high-frequency
parts of an image and suppresses the noises that exist in vein
pattern, Median filter which could reduce salt and pepper
noise, eliminates blurs and make the borderline smooth.
Filtered vein image undergoes morphological operation
which removes pixels on the boundaries of vein pattern but
does not allow them to break apart. The pixels remaining
make up the image skeleton. Thinning removes pixels so
that vein pattern without holes shrinks to a minimally
connected stroke, and the vein pattern with holes shrinks to
a connected ring halfway between each hole and the outer
boundary. The final image obtained after the pre-processing
stage is thinned and skeletonised image. Hough Transform
is used for feature extraction. Hough Transform can detect
curves. Prasanna et al. (2012) used unsharp high boost
filtering followed by histogram equalisation. Vein patterns
can also be obtained by image binarisation, edge detection,
key-points locating ROI extraction and ROI normalisation
(Bu et al., 2012).
The key objective while segmenting the ROI is to
automatically normalise the region in such a way that the
image variations caused by the interaction of the user with
the imaging device can be minimised. To make the
identification process more effective and efficient, it is
necessary to construct a coordinate system that is invariant/
robust to such variations. It is judicious to associate the
coordinate system to the palm itself since we are seeking the
invariance corresponding to it. Therefore, two webs
are utilised as the reference points/lines to build up the
coordinate system (Zhou and Kumar, 2011).
In our proposed ROI extraction the grey-scale image is
binarised to separate the palm region from the background
by regular thresholding. The threshold is deduced after
observing all the captured images. The web points are
observed to follow a specific pattern as follows.
The binary image is then scanned over a particular region to
locate the webs. The region is determined by analysing
all the images. It is observed that a cluster of points in a
single web follow this pattern so we eliminate the redundant
points so as to obtain a single point for each web.
A maximum of three webs are located by this algorithm.
Then the farthest two webs are used for further operations,
i.e., forming a square. This square is then projected
on the grey-scale image and then ROI is extracted after
rotating the square. The sides of the square are equal to the
distance between the two webs, which varies from person
to person so we resize all the extracted ROIs to a specific
measure.
As suggested by Zhou and Kumar (2011) the
background intensity profile of the extracted ROIs is
estimated by dividing the ROIs into overlapping blocks
of 8 × 8 pixels and the average grey-level in each block is
calculated. The blocks are overlapped by 3 pixels each.
This background intensity profile is then resized to the
original ROI and then subtracted from the ROI. Histogram
equalisation is employed to further enhance the image.
Processed ROIs are stored and used for matching. The entire
process described above is shown in Figure 4. The original
and enhanced ROI is shown in Figure 5.

Palm vein pattern-based biometric recognition system 109
Figure 4 (a) Binarised image; (b) ROI and (c) extracted ROI
(see online version for colours)
(a) (b) (c)
Figure 5 Enhanced ROI
3.2 Feature extraction and matching
Various sophisticated feature extraction and matching
techniques have been employed till date. Neighbourhood
Radon Transform is widely used. Texture analysis and
segmentation. Hamming distance is the parameter used for
matching. Linear hausdorff distance (LHD) between a pair
of vein patterns is also used. Fischer et al. (2012) have
suggested using the enhanced local binary patterns
histogram sequence (ELGBPHS) algorithm. Yuan and Li
(2012) have extracted features using affine invariant.
Spatial information and chain codes have been used
by Pflug et al. (2012) to extract features from vein images.
Sharavanan and Nagappan (2013) propose palm vein
authentication by using junction point with correlation
method.
We have aimed at devising simpler but accurate
matching techniques.
3.3 Principle component analysis
Kumari et al. (2011) have defined a good pattern
matching algorithm. We have used the same algorithm for
matching the veins. We have used our own ROIs of size
158 × 158 pixels. The PCA analysis of CASIA database
images have been also carried out.
PCA algorithm:
1 Read all ROIs and store its pixel values into array
forming a matrix having columns equal to total number
of ROIs read (Matrix A).
2 Calculate mean of this matrix (mean of each row).
1
(, )
n
j
M
iAij
=
=∑
3 Find deviation matrix from its mean matrix.
(, ) (, )Bi j Ai j Mi=−
4 Calculate eigen vectors of this deviation matrix
(Ev is Eigen vector).
EBEv=×
5 Project each image into Eigen face of deviation matrix.
Pj E Aj
′
=×
6 Read image to be matched and convert into grey-scale.
7 Extract ROI from image to be matched and enhance
it (T).
8 Form a vector with number of rows equal to size of test
ROI (Z).
9 Calculate difference matrix from mean matrix.
DZMi=−
10 Find projected test image.
QED
′
=×
11 Find minimum difference with projected test image.
1
1
_
|(,) (,)|
n
i
Euc dist P i j Q i j
n=
=−
∑
12 Image with minimum difference will represent the test
image.
3.4 2D discrete wavelet transform
We perform single level 2D Wavelet decomposition on the
processed ROIs (Polikar, 1996). The approximation and the
detailed coefficients matrices (horizontal, vertical and
diagonal, respectively) are computed. Further the energy,
mean, standard deviation, skewness and kurtosis of these
matrices is calculated. The computed parameters of each
image are stored in an array. All the arrays are then merged
to form a matrix in which each row corresponds to one
processed ROI. ROI is then extracted from the test image.
Pre-processing and single level 2D wavelet transform is
performed. As explained above the statistical parameters are
extracted from the processed test ROI and stored in an
array.
Euclidean distance is then computed between this row
array and each row of the matrix of statistical parameters.
Image corresponding to the row which has the minimum
distance is then adjudged as the matched image. The
wavelet used for the decomposition is ‘db4’.
3.5 Template matching
In our proposed template matching technique all the
processed ROIs are reshaped to a row vector. These vectors
are then stored in a matrix such that each row corresponds
to one image. ROI is extracted from the test image and
enhanced. It is also reshaped to a row vector.
The difference is computed between the test ROI vector
and each row of the matrix. We get a difference vector.

110 G. Shah et al.
All the values in the difference vector are summed.
Image corresponding to the row which has minimum sum is
then adjudged as the matched image.
4 Experimental results
Experimentation was carried using MATLAB 7.2 with
Image Processing toolbox on i3 machine with 4 GB RAM.
Palm vein images of 62 persons were used for
experimentation. The images used for testing are taken at
different times. The original (captured at first time) is used
to generate template/feature. The accuracy for each
technique is as shown in Table 1. Better accuracy can be
acquired if the images are acquired with uniform
illumination throughout. This setup can be modified to use
an IR sensitive camera to capture better images. Also the
design of the setup can be made more compact to keep the
atmospheric conditions inside it constant so that they do not
hamper the quality of the images.
Table 1 Results of three methods used for recognition
Methods % accuracy
PCA 85.48
2D wavelet transform 39
Template matching 93.54
5 Conclusion
The experimental set up designed to capture palm vein
images is cost effective and simple to develop. It consists of
simple components which include 850 nm IR LEDs to
illuminate the palm and a low-cost webcam to capture the
images at a specific distance. The currently available
systems in the market cost around $4000. Processing is done
only on the ROI which saves memory and also increases
the speed of processing. The extraction of the ROI from the
image of the palm makes the matching technique rotation
invariant. The images of the palm veins obtained by our
experimental setup have a comparatively good contrast and
the veins can be clearly distinguished. These images are by
far the best available images. Template matching has been
proved to be the best method compared with PCA and 2D
wavelet transform with an accuracy of 93.54%. Being the
intensity variations in the images are small, wavelet
statistics do not provide good results. However, the images
can be denoised and decomposed up to three levels and
feature vector can be prepared. The highest accuracy
recorded so far is 99% using 2D Gabor filter (Lee, 2012).
It was found difficult to work on the same database with our
proposed technique.
Palm veins can also be fused with other biometric
modalities to get more accurate identification/recognition
results. The palm veins can be used as multiple distinctive
units (left and right hands which have different patterns).
The palm veins if fused with palm print, face, iris or finger
print recognition can give us best results. The veins cannot
be forged from a surface like the finger prints nor are they
affected by diseases like the iris recognition. Fusion of
various modalities gives us more parameters for testing.
Each modality can compensate for the shortcoming of the
others hence reducing the chances of false recognition and
failure.
References
Bashkatov, A.N., Genina, E.A., Kochubey, V.I. and Tuchin, V.V.
(2005) ‘Optical properties of human skin, subcutaneous and
mucous tissues in the wavelength range from 400 to 2000
nm’, Journal of Physics D: Applied Physics, Vol. 38,
No. 15, p.2543.
Bhattacharyya, D., Ranjan, R., Farkhod Alisherov, A. and
Choi, M. (2009) ‘Biometric authentication: a review’,
International Journal of u-and e-Service, Science and
Technology, Vol. 2, No. 3, pp.13–28.
Bu, W., Wu, X. and Gao, E. (2012) ‘Hand vein recognition
based on orientation of LBP’, SPIE Defense, Security, and
Sensing, International Society for Optics and Photonics,
pp.83711Y–83711Y.
Fischer, M., Rybnicek, M. and Tjoa, S. (2012) ‘A novel palm vein
recognition approach based on enhanced local gabor binary
patterns histogram sequence’, Proceedings of 19th IEEE
International Conference on. In Systems, Signals and Image
Processing (IWSSIP), 2012, pp.429–432.
Ghosh, P. and Dutta, R. (2012) ‘A new approach towards
biometric authentication system in palm vein domain’,
International Journal of Advance Innovations, Thoughts &
Ideas, Vol. 1, No. 2, pp.1–10.
Han, W.Y. and Lee, J.C. (2012) ‘Palm vein recognition using
adaptive Gabor filter’, Expert Systems with Applications,
Vol. 39, No. 18, pp.13225–13234.
Jia, X., Cui, J., Xue, D. and Pan, F. (2012) ‘A adaptive dorsal hand
vein recognition algorithm based on optimized HMM’,
Journal of Computational Information Systems, Vol. 8, No. 1,
pp.313–322.
Kumar, A. and Prathyusha, K.V. (2008) ‘Personal authentication
using hand vein triangulation’, Proceedings of the SPIE
Biometric Technology for Human Identification, Vol. 6944,
pp.69440E–69440E.
Kumari, L., Debbarma, S., Debbarma, N. and Deb, S. (2011)
‘Image pattern matching using principal component analysis
method’, International Journal of Advanced Engineering and
Application, Vol. 1, pp.6–9.
Lee, J.C. (2012) ‘A novel biometric system based on palm
vein image’, Pattern Recognition Letters, Vol. 33, No. 12,
pp.1520–1528.
Nandini, C., Ashwini, C., Aparna, M., Ramani, N. and Kini, P.
(2012) ‘Biometric authentication by dorsal hand vein pattern’,
International Journal of Engineering and Technology, Vol. 2,
No. 5, pp.837–840.
Pflug, A., Hartung, D. and Busch, C. (2012) ‘Feature extraction
from vein images using spatial information and chain codes’,
Information Security Technical Report, Vol. 17, No. 1,
pp.26–35.
Polikar, R. (1996) The Wavelet Tutorial, http://users.rowan.edu/
~polikar/WAVELETS/WTpart1.html
Prasanna, R.D., Neelamegam, P., Sriram, S. and Raju, N. (2012).
‘Enhancement of vein patterns in hand image for biometric
and biomedical application using various image enhancement
techniques’, Procedia Engineering, Vol. 38, pp.1174–1185.

Palm vein pattern-based biometric recognition system 111
Sarkar, I.S., Alisherov, F.A., Bhattacharyya, D.B. and Kim, T.H.K.
(2010) ‘Palm vein authentication system: a review’,
International Journal of Control and Automation, Vol. 3,
No. 1, pp.27–34.
Sharavanan, S. and Nagappan, A. (2013) ‘A hand palm vein
authentication by using junction points with correlation
method’, International Journal of Computational
Engineering Research, Vol. 3, No. 1, pp.77–84.
Wang, L. and Leedham, G. (2006) ‘Near-and far-infrared
imaging for vein pattern biometrics’, Proceeding of IEEE
International Conference on Video and Signal Based
Surveillance, AVSS’06. pp.52-52.
Wang, L., Leedham, G. and Cho, S.Y. (2007) ‘Infrared imaging of
hand vein patterns for biometric purposes’, IET Computer
Vision, Vol. 1, No. 3, pp.113–122.
Yuan, W.Q. and Li, W. (2012) ‘A palm vein feature extraction
method based on affine invariant’, Proceedings of IEEE
International Conference on Robotics and Biomimetics
(ROBIO), 2012, pp.2323–2326.
Zhang, D. (2000) Automated Biometrics: Technologies and
Systems, Springer.
Zhou, Y. and Kumar, A. (2011) ‘Human identification using palm-
vein images’, IEEE Transactions on Information Forensics
and Security, Vol. 6, No. 4, pp.1259–1274.
Website
CASIA Database, http://biometrics.idealtest.org/dbDetailForUser.
do?id=6