Professional Documents
Culture Documents
2012 - Elsevier - A Novel Biometric System Based On Palm Vein Image
2012 - Elsevier - A Novel Biometric System Based On Palm Vein Image
a r t i c l e i n f o a b s t r a c t
Article history: Vein pattern recognition is one of the newest biometric techniques researched today. In this paper, one of
Received 15 July 2011 the reliable and robust personal identification authentication approaches using palm vein patterns is pre-
Available online 25 April 2012 sented. We consider the palm vein as a piece of texture and apply texture-based feature extraction tech-
niques to palm vein authentication in our work. A 2-D Gabor filter provides the optimized resolution in
Communicated by A. Heyden
both the spatial and frequency domains, thus it is a basis for extracting local features in the palm vein
recognition. In order to obtain effective pattern of palm vascular, we proposed an innovative and robust
Keywords:
directional coding technique to encode the palm vein features in bit string representation. The bit string
Biometrics
Palm vein recognition
representation, called VeinCode, offers speedy template matching and enables more effective template
Gabor filter storage and retrieval. The similarity of two VeinCodes is measured by normalized hamming distance.
Feature extraction A total of 4140 palm vein images were collected form 207 persons to verify the validity of the proposed
palm vein recognition approach. High accuracy has been obtained by the proposed method and the speed
of the method is rapid enough for real-time palm vein recognition. Experimental results demonstrate that
our proposed approach is feasible and effective for palm vein recognition.
Ó 2012 Elsevier B.V. All rights reserved.
0167-8655/$ - see front matter Ó 2012 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.patrec.2012.04.007
J.-C. Lee / Pattern Recognition Letters 33 (2012) 1520–1528 1521
such as the geometry, the principal line, and the delta point. Fur- 1994), fingerprint recognition (Lee and Wang, 1999), face recogni-
thermore, hand veins show significant textural differences and a tion (Liu, 2004), palmprint recognition (Zhang et al., 2003), iris rec-
lot of minutiae similar to the ridges and branches of fingerprints. ognition (Daugman, 1993) and so on. Because the palm vein image
These features offer stable, unique and reliable biometrics for per- consists of many lines, it has stable and obvious direction which
sonal identification. Many verification technologies using biomet- makes it suitable to be viewed as a texture image. In this work,
ric features of hand veins have been developed over the past the 2-D Gabor filter based palm vein recognition method is de-
decade. Lin and Fan (2004) present personal verification results scribed in detail. Experimental results on our own NIR palm vein
using palm dorsal images acquired from a thermal infrared (IR) image database, including more than 200 classes, have demon-
camera operating in the 3.4–5.0 lm range. Their approach is based strated that the proposed method is highly feasible and effective
on the combination of multi-resolution images obtained from the for personal identification. The palm vein identification system
pre-processed thermal vein images. Wang et al. (2007) present an- based on 2-D Gabor filter consists of two parts: a palm vein collec-
other approach for personal authentication using hand vein images tion device for palm vein image acquisition and an algorithm for
acquired from the thermal imaging. The authors (Wang et al., 2007, real-time palm vein identification. There are three main issues to
2008a) have employed Hausdorff distance to generate matching be considered in the system:
scores between the extracted line patterns and illustrated promis-
ing results. Crisan et al. (2010) focused on improving the two (1) The palm vein images are captured by a NIR camera as the
essential parts of a vein-scanning device: the hardware lighting input data. Then, a median filter is employed on the palm
system and the feature extraction algorithms. Hsu et al. (2011) vein images to remove some noises and an Otsu’s 1979
has been successfully used the modified 2-directional 2-dimen- method is applied to select a suitable threshold to segment
sional principal component analysis ((2D)2PCA) to obtain eigenve- the palm region. Finally, two baseline data points are found
ins which is a low dimensional representation of dorsal hand vein automatically and a rectangular region of interest (ROI) is
patterns. However, there are rarely published papers about the obtained.
palm vein recognition. Most of the papers are white papers (2) 2-D Gabor filter was used on the palm vein pattern to extract
published by Fujitsu Company (Sasaki et al., 2005, and Fujitsu features. To obtain useful representation of the palm vein
Laboratories Ltd, 2003). Fujitsu’s palm vein verification product modalities, we propose a new scheme named directional
has high accuracy, but to the best of our knowledge, the features coding method to encode the palm vein features in bit string
used are not disclosed in any published research articles. representation.
In the field of the palm vein recognition, some papers with differ- (3) The Hamming distance is employed to measure the similar-
ent technologies are described below. Wang et al. (2008b) proposed ity of two VeinCodes and verify whether the template and
a multimodal personal identification system where palmprint and testing samples are captured from the same person or not.
palm vein modalities were combined in a single image. Locality Pre-
serving Projection (LPP) was used to extract features of the fused The rest of this paper is organized as follows. We will briefly
images, which they called ‘‘Laplacianpalm’’. Zhou and Kumar introduce the palm vein image collection device and process in
(2011) employed the Hessian phase approach to feature extraction Section 2. A detailed description of the proposed method for palm
and presented a matching approach that can effectively accommo- vein recognition is given in Section 3. Experimental results are pre-
date the potential image deformations, translational and rotational sented and discussed in Section 4, prior to conclusions in Section 5.
variations by matching to the neighborhood of the corresponding
regions and generate more reliable matching scores. Zhang et al.
(2007) used a low cost CCD camera to capture the infrared palm 2. Palm vein images collection
images and extracted the palm vein features by multi-scale filtering.
The experimental results demonstrate that the recognition rate of In visible light, the vein structure of the palm is not always eas-
their system is fine but not good enough to be a real system. Michael ily discernible. The visibility of the vein structure varies signifi-
et al. (2010) presented an innovative contactless palm print and cantly depending on such factors as age, level of subcutaneous
palm vein recognition system. The proposed system offers several fat, ambient temperature and humidity and physical activity. In
advantages like low-cost, accuracy, flexibility, and user-friendli- addition a multitude of other factors including scars, pigmentation
ness. Although these authors claim the features they use to recog- and palm position can also effect upon the image acquisition. Up to
nize the vein patterns can attain a high level of accuracy, most of now, there is no publicly available palm vein pattern database
them used small datasets to evaluate the performance of hand vein available to the research community. In order to perform a preli-
recognition, so the practical feasibility of these algorithms has not minary analysis on the features of the palm vein pattern, we con-
been demonstrated. structed our own NIR palm vein image database. In this work, a
Up to now, there is currently no publicly available palm vein low-cost NIR charge-coupled device (CCD) camera that we used
pattern database provided to the research community. To demon- to acquire palm vein images was a digital noise reduction (DNR)
strate the robustness of the proposed approach, we constructed DSP Camera. It possesses the benefits of high availability, uniform
our own near-infrared (NIR) palm vein pattern image database, and consistent good image quality, convenience and low cost.
which has 207 distinct participants and contains 4140 palm vein Though principally designed for use in visible light, CCD cameras
pattern images. It is based on a contactless, non-invasive data are also sensitive to NIR wavelengths of the electromagnetic spec-
acquisition method and requires no injection of any agents into trum up to about 1100 nm. This is within the actinic IR range,
the blood vessels. Therefore, it is by far the best-known non-inva- which covers the NIR spectrum from 700–1400 nm. This wave-
sive for acquiring palm vein pattern images. length also lies in the medical spectral window (700–900 nm) in
Gabor transform is first brought forward by Dennis Gabor in which the illumination penetrates deeper into the biological tis-
1946. Daugman (1980) explains how 2-D Gabor filter can attain sues (Godik and Guljaev, 1991). In our work, the near IR illumina-
the combined optimized resolution in space and frequency. tion (LEDs) is evenly and circularly located around the camera,
Namely, it can adjust the direction and frequency band width producing a peak at 750 nm wavelength. In addition to the LEDs,
when doing space localization. This characteristic is very useful another important part of this system is an infrared filter. To elim-
for analyzing the texture of images. The method has been success- inate the effect of visible light, an optical infrared filter was
fully used in texture segmentation and classification (Dunn et al., mounted in front of the camera’s lens. For our data collection, a
1522 J.-C. Lee / Pattern Recognition Letters 33 (2012) 1520–1528
Hoya RM80 IR filter was used. Since there are no docking devices to rithm (Sonka et al., 1999) is employed to find the palm border.
constrain the participant’s hand, it can be used for either the right Fig. 1(c) shown an example illustrating how the extracted border
or left hand. of a palm image which perfectly matches the original palm con-
Using the NIR light source and the IR filter, the image on the tour. Then for each point on the palm contour, the distance be-
CCD sensor is a high quality image of palm vein. The captured palm tween this point and the mid-point of the wrist is calculated.
vein images are 8-bit gray images with a resolution of 320 240. Fig. 1f shows the distance distribution diagram. As can be seen,
The reason for using such a low spatial resolution is that both there are five local maximums and four local minimums. The pat-
the data amount in palm vein images and noise sensitivity can tern in the diagram is quite similar to the geometric shape of a
be reduced while the features of palm veins can still be preserved. palm (see Fig. 1(a)), which also has five tips (local maximums)
During the collection of images for the database, we restricted the and four finger-webs (local minimums). Experimenting on a wide
position where the participants placed their palms above the cam- variety of palm vein images, we found that the four local minimum
era. The distance between the hand and the camera was adjusted locations in the distance distribution diagram are the same as fin-
to approximately 20 cm for the best acquisition. At this distance, ger-web locations and match between the two locations is very
there is acceptable tolerance for positioning the palm within the close. Finally, two valley points, p1 (the valley point between the
specified region. In our palm vein database, 207 volunteers were small finger and ring finger) and p2 (the valley point between the
enrolled in our study and palm vein images were captured from middle finger and the index finger), are selected as two key data
their left hands at two different times after an interval of at least points, as shown in Fig. 1(d). These two data points (p1 and p2;)
one month. Ten palm vein images were acquired each time for are employed to locate the ROI. The procedure is described as
each person. The database constructed in this way includes possi- follows.
ble variations of palm vein images under various conditions. It also First, the straight line p1 p2 is formed by the points of p1 and p2
includes several racial groups, including Chinese, African and Cau- as shown in Fig. 1d. To eliminate the influences of palm rotation
casian. The age range of the volunteers was between 18 and and define the coordinates of ROI more conveniently, the palm im-
60 years, and their occupations ranged from university students, age is rotated by the angle h (Eq. (1)) between line p1 p2 and the
professors and technicians to manual workers such as cleaners horizontal line.
and electricians. No distinction was made between male and fe-
h ¼ tan1 ðY p2 Y p1 Þ=ðX p2 X p1 Þ; ð1Þ
male when the samples were collected. When seen on the com-
puter monitor, the mostly distinguishable component in the where ðX p1 ; Y p1 Þ is the coordinate of p1 and (X p2 ; Y p2 ) is the coordi-
database image is the superficial vein tree pattern. In our study, nate of p2.
the quality of the palm vein images does not change significantly This makes the direction of line p1 p2 is horizontal, as shown in
in an air-conditioned or outdoor environment. Fig. 1(e). A rectangular region Rc1 c2 c3 c4 whose corners are c1, c2, c3
and c4 is located and denoted as the ROI. The length of line c1 c2
3. Proposed method for palm vein recognition and linepffiffiffip1 p2 are equal. The Line c1 c2 and c1 c3 are vertical and
c1 c3 ¼ 2 c1 c2 . Finally, we redefine the original coordinate
In this paper, a palm vein recognition algorithm is presented, (0, 0) at the upper-left point c1 of the ROI. The reason for extracting
which includes three basic processes: preprocessing, feature the ROI in this manner is that this ensures all the ROIs reference
extraction and matching. Detailed descriptions of these three steps the same region in the palm vein image. In this way, we can over-
are introduced in the following sections. come the problem of not using docking devices while acquiring
palm vein images. High verification accuracy can simultaneously
be maintained.
3.1. Preprocessing and extraction of the ROI
3.2. Palm vein feature extraction by texture analysis
Image segmentation is one of the most important steps in the
analysis of processed image data. Its main goal is to separate com- This section defines our palm vein feature extraction method,
ponents of the object image that have a strong relevance to the which includes filtering and matching. The motivation for using a
analysis from the background. When a palm image is obtained, Gabor filter in our palm vein research is first discussed.
the palm background is first segmented from the image. Binariza-
tion is used to segmenting the image into two levels; object (palm 3.2.1. Overview of Gabor functions
region) and background. For palm image segmentation, Otsu’s Marcelja (1980) and Daugman (1980, 1985) modeled the re-
thresholding is applied to the palm image to estimate the palm sponses of the visual cortex by Gabor functions because they are
region. By comparing the segmented palm shapes with the of ori- similar to the receptive field profiles in the mammalian cortical
ginal palm images, we can see that they are almost indistinguish- simple cells. Gabor filters are extremely useful for texture analysis
able as shown in Fig. 1(a) and (b). This shows that Otsu’s method is because of the 2-D spectral specificity of texture as well as its var-
quite effective in determining the threshold of palm images. iation with 2-D spatial position. Daugman (1980, 1985) developed
To increase verification accuracy and reliability, the features of the 2D Gabor functions (a series of local spatial bandpass filters),
vein patterns extracted from the same region in different palm vein which have good spatial localization, orientation selectivity, and
images are compared for verification. The region to be extracted is frequency selectivity. In addition to accurate 2D space and 2D spa-
known as the ROI. For this reason, it is important to fix the ROI in tial frequency location, they also provide robustness against vary-
the same position in different palm vein images to ensure the sta- ing brightness and contrast of images (Zhang et al., 2003). 2-D
bility of the principal extracted vein features. However, it is diffi- Gabor filter has the following general form:
cult to fix the ROI at the same position in different palm vein ( !)
2 2
images without the use of a docking device to constrain the palm 1 xh k y h k
position when acquiring palm vein images. In this paper, we have hðx; y; hk ; f ; rx ; ry Þ ¼ exp þ 2 expði2pfxhk Þ; ð2Þ
2 rx ry
2
employed the two data points (p1 and p2, as shown in Fig. 2(d)), to
determine the approximate (not absolute) immovable ROI. The fol- In Eq. (2), xhk ¼ x cos hk þ y sin hk and yhk ¼ x sin hk þ y cos hk , f is the
lowing processes are performed to determine the two data points frequency of the sinusoidal plane wave, hk is the orientation of
in binary palm vein images. First, the inner border tracing algo- the Gabor filter, and rx and ry are the standard deviations of
J.-C. Lee / Pattern Recognition Letters 33 (2012) 1520–1528 1523
Fig. 2. Example of directional code being applied on palm vein image. (a) is the original palm vein image, (b) to (e) show the imaginary part of the palm vein in the four
directions (h = 0, 45, 90, 135), and (f) is the result of taking the maximum gradient values obtained from the four responses.
1524 J.-C. Lee / Pattern Recognition Letters 33 (2012) 1520–1528
the Gussian envelope along the x -axis and y -axis, respectively. w=21
X w=21
X
To analysis the Gabor filter in terms of the real part and imag- g Im ðX; Y; hk ; f ; rx ; ry Þ ¼ IðX þ x; Y þ yÞ
x¼w=2 y¼w=2
inary part, we express Eq. (2) in the complex form h = hRe + ihIm
where hIm ðx; y; hk ; f ; rx ; ry Þ ð6Þ
( !)
2 2
1 xhk yhk where I(,) denotes a w w 256 gray-level image.
hRe ðx; y; hk ; f ; rx ; ry Þ ¼ exp þ 2 cosð2pfxhk Þ; ð3Þ In the proposed method, we proposed a new scheme named
2 rx ry
2
directional coding method to extraction the maximum gradient
( !) values of the palm vein from the different orientation. Four filters
2 2
1 xhk yhk which have same frequency and variance but different directions
hIm ðx; y; hk ; f ; rx ; ry Þ ¼ exp þ 2 sinð2pfxhk Þ; ð4Þ
2 rx ry
2 are chosen to extract direction information. The direction of point
(x, y) is defined as:
Such Gabor filters are widely used as tunable filters for extract- dRe ðx; yÞ ¼ arg max fg Re ðX; Y; hk ; f ; rx ; ry Þg ð7Þ
k¼1;2;3;4
ing the orientation or edge information from images. Lee (2003)
gave a good introduction to image representation by using Ga-
bor functions. To obtain high efficiency and better resolution dIm ðx; yÞ ¼ arg max fg Im ðX; Y; hk ; f ; rx ; ry Þg ð8Þ
k¼1;2;3;4
in both space and frequency domains, Daugman (1993) applied
the 2-D Gabor filter on iris recognition. Based on these proper- Fig. 2(b)–(e) shows the imaginary part of the palm vein in the four
ties, in this paper, we try to apply a Gabor filter to palm vein directions (h = 0, 45, 90, 135).Fig. 2(f) is the result of taking the max-
authentication. imum gradient values obtained from the four responses. This image
depicts the strongest directional response of the palm vein and it
3.2.2. Filtering and feature extraction closely resembles the original palm vein pattern shown in Fig. 2a.
Palm vein refers to the vascular pattern or blood vein patterns Finally, the each pixel in the maximum gradient image is coded to
recorded from underneath the human skin. Every person has un- two bits, called VeinCode (Vr, Vi), by the following inequalities.
ique structure and position of veins, and this does not change sig- V r ðx; yÞ ¼ 1 if dRe ðx; yÞ P 0; ð9Þ
nificantly from the age of ten (Biometric Technology Today,
2004). Because the palm veins contains vascular network which V r ðx; yÞ ¼ 0 if dRe ðx; yÞ < 0; ð10Þ
also resembles line-like characteristic, it has stable and obvious
direction which makes it suitable to be viewed as a texture im- V i ðx; yÞ ¼ 1 if dIm ðx; yÞ P 0; ð11Þ
age. Even though the line-based approach plays an important role
in palm vein authentication, it is difficult to extract them accu- V i ðx; yÞ ¼ 0 if dIm ðx; yÞ < 0; ð12Þ
rately in practice. 2-D Gabor filter is widely used in extracting
the features of textured images. This motivates us to apply 2-D Using this directional coding method, the direction information in
Gabor filter to palm vein authentication. Considering that palm palm vein images is stored in the feature vector. Fig. 3d and e show
vein also has obvious direction and local features, our palm vein the real part feature of two palm veins belonging to the same per-
identification method uses 2-D Gabor filters with four different son with m = 4 (h = 0, 45, 90, 135). Fig. 3(d) and (f) show the real part
orientations to filter the original ROI image. Then the direction feature of two palm veins belonging to the different persons. From
texture features are extracted. The original palm vein feature Fig. 3, we find that the same persons real Gabor features are similar
information is to a large extent preserved. The proposed direc- and those of different persons are not. This reveals that the feature
tional coding technique aims to encode the line pattern based based on 2-D Gabor filter can be used as the palm vein feature.
on the proximal orientation of the lines.
Since most local palm vein structures come with a well-de- 3.2.3. Palm vein matching
fined local frequency and orientation, we set f as the reciprocal In order to describe clearly the matching process, each feature
of the average inter-vein distance and m as the number of ori- vector is considered as two 2-D feature matrices, real and imagi-
entations for calculating hk = p(k 1)/m, k = 1, . . ., m. After decid- nary. A normalized hamming distance (Daugman, 1993) is adopted
ing the parameters of the 2-D Gabor filters, the real and to determine the similarity measurement for VeinCode matching.
imaginary parts at the sampling point (X, Y) can be defined as Let P and Q be two palm vein feature matrices. The normalized
follow: hamming distance can be defined as
PminðN;NþuÞ PminðM;Mþv Þ
i¼maxð1;1þuÞ j¼maxð1;1þv Þ ðP R ði þ u; j þ v Þ Q R ði; jÞ þ PI ði þ u; j þ v Þ Q I ði; jÞÞ
HDmin ¼ min ; ð13Þ
juj<U;jv j<V 2HðuÞHðv Þ
w=21
X w=21
X where PR(QR) and PI(QI) are the real part and the imaginary part of P(Q),
g Re ðX; Y; hk ; f ; rx ; ry Þ ¼ IðX þ x; Y þ yÞ respectively; the Boolean operator, ‘‘’’, denotes Exclusive-OR, the size
x¼w=2 y¼w=2
of the feature matrices is N M, the U = 3, and V = 3 determine the max-
hRe ðx; y; hk ; f ; rx ; ry Þ ð5Þ imum allowable horizontal and vertical translation, H(u) = min (N,-
N + u) - max (1,1 + u), and H(v) = min (M,M + v) - max (1,1 + v). It is
J.-C. Lee / Pattern Recognition Letters 33 (2012) 1520–1528 1525
Fig. 3. (a)–(c) show the three original palm vein images. (d)–(f) show the real part Gabor features of the (a)–(c) palm vein images.
noted that HDmin is between 1 and 0. The hamming distance for perfect to analyze the relationship between the recognition accuracy and
matching is zero. Although we employed two data points to find the ROI the different resolutions of ROI images. The second one is to iden-
automatically in the preprocessing, there are still a few translations and tify the number of orientations hk selected for the proposed meth-
rotations during the location of ROI. In order to provide translation od and to further evaluate the real and imaginary Gabor features
invariance and rotation matching, the minimum HD value obtained for palm vein recognition, respectively. The following subsections
from the Eq. (13) which is considered to be the final matching score. detail the experiments and results.
At present, only small data sets are used to evaluate recognition During the experiments, each ROI image is decomposed into
performance for palm vein recognition by most algorithms. Only different levels (120 169, 60 84, 30 42 and 15 21 resolu-
the approaches proposed by Zhou and Kumar (2011) and Chen tions) and the images at each level are been tested. Therefore, there
et al. (2009) have been tested on large image sets involving more are four levels, the image with 120 169 resolution is the 1st level
than 200 subjects. To demonstrate the robustness of the proposed and 15 21 resolution is the 4th level. The well-known statistical
approach, in our experiments, the total number of palm vein pair, the false rejection rate (FRR) and the false acceptance rate
images was 4140, which were collected from 207 persons each (FAR), is usually adapted to analysis evaluate the performance of
with 20 palm vein images captured at two different times. The size the experimental results. Figs 4(a)–(d) illustrate the FAR and FRR
of each palm vein image was 320 240 resolution and 256 gray- distributions. It can be seen from those figures that there exists
levels. The experiments are conducted in two modes: identification two curves in the distributions at different resolution. One curve
(i.e., one-to-many matching) and verification (i.e., one-to-one corresponds to genuine matching and the other corresponds to
matching). In identification mode, the algorithm is measured by impostor matching. When the decision threshold is set as the
correct recognition rate (CRR), the ratio of the number of samples one corresponding to the intersection of genuine and impostor dis-
being correctly classified to the total number of test samples. In tribution curves, the total error reaches the minimum, and the cor-
verification mode, the receiver operating characteristic (ROC) responding threshold. According to the Fig. 4, equal error rates
curve is used to report the performance of the proposed method. (EERs) of the 3rd levels are much less than those of other levels.
Four kinds of experiment schemes were designed as follow: four In other words, the palm vein image with 30 42 resolution is
(six, eight or ten) images of each person were randomly selected more suitable for feature extraction than the other resolutions. In
for training, and the other ten samples were used as the testing Table 1, it is evident that the image resolution can play an impor-
images to verify the validity of the proposed approach. In this sec- tant role in the matching process. Therefore, the image with higher
tion, two experiments conducted to demonstrate the performance or lower resolution is not suitable for the method proposed in this
of the proposed method for palm vein recognition. The first test is paper, and they can not offer higher recognition rate. High resolu-
1526 J.-C. Lee / Pattern Recognition Letters 33 (2012) 1520–1528
Density
Density
0.1 0.1
0.05 0.05
0 0
0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.3 0.4 0.5 0.6 0.7 0.8 0.9
Matching Score Matching Score
Density
0.1 0.1
0.05 0.05
0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0 0.1 0.2 0.3 0.4 0.5 0.6
Matching Score Matching Score
Fig. 4. FAR and FRR distributions at each level: (a) 1st level, (b) 2nd level, (c) 3rd level, and (d) 4th level.
Table 2
The recognition rate of proposed methods obtained by tested 2-D Gabor with 4.2. Comparison and discussions
different orientations.
Methods Number of different orientations Among existing methods for palm vein recognition, those pro-
posed by Wang et al. (2008a), Wang et al. (2008b), and, Zhou and
m = 4 (%) m = 8 (%)
Kumar (2011), respectively, are the best-known. Moreover, they
Only Real 94.64 95.67
characterize local details of the palm vein from different viewpoints,
Only Imaginary 91.47 93.15
Real + Imaginary 99.18 99.35 i.e. the geometric-based approach, appearance-based approach, and
phase based representation method. To further prove the effective-
ness of the proposed method, we make detailed comparison be-
tion images are mixed with noise and low resolution images lose tween the proposed method and the above three methods (and
too much information. The relationship between the verification our previous work (Hsu et al., 2011)) on our own palm vein data-
result and the different number of training samples are also shown base. We compare our experimental results of each algorithm with
as Table 1. It reveals that as the number of training samples per its published results and find that they are consistent. This verifies
class is increased and the relative gain among those resolutions be- to a large extent the correctness of our implementations of other
comes more apparent. methods. Table 3 and Fig. 5 detail the experimental results.
In this paper, we proposed the directional coding method to ex- Table 3 illustrates that the CRR of each algorithm is greater than
tract maximum gradient values of the palm vein by combining the 90%. This implies the high accuracyof these methods. Fig. 5 illus-
real part with imaginary part Gabor features. However, there are trates the ROC curves on our palm vein database for the five algo-
more than four orientations to select in the 2-D Gabor filter. There- rithms compared, namely, laplacianpalm, minutiae feature points,
fore, in second test, we tested 2-D Gabor with different orienta- Hessian phase, eigenvein and the proposed method. To better dis-
tions, and then applied the proposed method to extract the tinguish these well performing methods, the horizontal axis of
maximum gradient palm vein image. The result of recognition Fig. 5 is spread out using the logarithmic scale. From the results
rates with no rejection option for m = 4 and m = 8 are shown in Ta- shown in Table 3 and Fig. 5, we can find that proposed method
J.-C. Lee / Pattern Recognition Letters 33 (2012) 1520–1528 1527
Table 3
Performance comparison of several well-known methods on our palm vein database. Laplacian Palm
0.07
Minutiae Feature
Methods Correct recognition Equal error Hessian Phase
rate (%) rate (%) 0.06 Eigenvein
Proposed
Minutiae feature points (Wang 97.68 1.84
et al., 2008a) 0.05
0.02
5. Conclusions Daugman, J.G., 1985. Uncertainty relation for resolution in space, spatial frequency
and orientation optimized by two-dimensional visual cortical filters. J. Optical
Soc. Am. 2 (7), 1160–1169.
We have developed a reliable and robust palm vein identifica- Dunn, D., Higgins, W.E., Wakeley, J., 1994. Texture segmentation using 2D gabor
tion system for real-time personal identification by applying a elementary functions. IEEE Trans. Pattern Anal. Mach Intell. 16 (2), 130–149.
Fujitsu Laboratories Ltd: Fujitsu Laboratories develops technology for world’s first
low-cost NIR CCD camera-based palm vein device to capture the
contactless palm vein pattern biometric authentication system, [Online March
palm vein images. A preprocessing algorithm extracts a rectangle 2003], available at: <http://pr.fujitsu.com/en/news/2003/03/31.html>.
area (ROI) from a palm vein image for feature extraction. To repre- Gabor, D., 1946. Theory of communication. J. Inst. Electr. Eng. 93, 429–457.
Godik, E.F., Guljaev, Y.V., 1991. Functional imaging of the human body. IEEE Trans.
sent a low-resolution palm vein image and match different palm
Eng. Med. Biol. 10, 21–29.
vein images, we extend the use of 2-D Gabor filter to represent a Hawkes, P.L., Clayden, D.O., 1993. Veincheck research for automatic identification of
palm vein image using its texture feature, and apply a normalized people. Presented at the Hand and Fingerprint Seminar at NPL. pp. 230–236.
hamming distance for the matching measurement. In addition, we Hsu C. B., Hao S. S., Lee J. C., 2011. Personal Authentication through Dorsal Hand
Vein Patterns, Optical Engineering, 47, 067205–1–067205-10.
proposed a new technique called directional coding to code the Jain, A.K., Bolle, R., Pankanti, S., 1999. Biometrics Personal Identification in
palm vein features in two bits representation. This method repre- Networked Society. Kluwer Academic Publishers, Massachusetts.
sents the biometric features in bit string format which enable spee- Lee, C.J., Wang, S.D., 1999. Fingerprint feature extraction using Gabor filters.
Electronic Lett. 35 (4), 288–290.
dy matching and convenient storage. Using this representation, the Lee, T.S., 2003. Image representation using 2D Gabor wavelets. IEEE Trans. Pattern
total size of a palm vein feature is reduced to 2520 bits. In our palm Anal. Mach. Intell. 18 (10), 959–971.
vein database of 4,140 palm vein images from 207 different palms, Lin, C.-L., Fan, K.-C., 2004. Biometric verification using thermal images of palm-
dorsa vein patterns. IEEE Trans. Circuits Syst. Video Technol. 14 (2), 199–213.
we can achieve high recognition rate (> 99%), and its equal error Liu, C., 2004. Gabor-Based Kernel PCA with fractional power polynomial models for
rate is 0.4%, which is comparable with all other hand-based bio- face recognition. IEEE Trans. Pattern Analy. Mach. Intell. 26 (5), 572–581.
metrics, such as hand geometry and fingerprint verification. MacGregor, P., Welford, R., 1991. Veincheck: imaging for security and personnel
identification. Adv. Imaging 6 (7), 52–56.
In summary, we conclude that our palm vein identification sys-
Masaki, W., Toshio, E., Morito, S., and Shigeru S., 2005. Palm vein authentication
tem can achieve good performance in terms of speed and accuracy. technology and its applications. In: The Biometric Consortium Conference. pp.
For further improvement of the system, we will focus on three is- 1–2.
Marcelja, S., 1980. Mathematical description of the responses of simple cortical
sues: (1) to reduce the size of the device for practical applications,
cells. J. Optical Soc. Am. 70 (11), 1297–1300.
(2) to design a hand attachment frame for constraining the partic- Michael, G. K. O., Tee, C., Jin, A. T. B., 2010. Design and implementation of a
ipant’s palm for stable preprocessing, and (3) to combine the pro- contactless palm print and palm vein sensor. In: 11th International Conference
posed palm vein coding scheme with other texture feature on Control, Automation, Robotics and Vision, pp.1268–1273.
Otsu, N., 1979. A threshold selection method from gray-level histograms. IEEE
measurements such as texture energy for coarse level classification Trans. Syst. Man Cybernet. 9 (1), 62–66.
to achieve higher performance. Vein recognition in Europe. 2004. Biometric Technology Today, 12(9), 6.
Wang, L., Leedham, G., Cho, D.S., 2007. Infrared imaging of hand vein patterns for
biometric purposes. Inst. Eng. Technol. Comput. Vision. 1, 113–122.
6. Uncited references Wang, L., Leedham, G., Cho, D.S., 2008a. Minutiae feature analysis for infrared hand
vein pattern biometrics. Pattern Recogn. 41 (3), 920–929.
Wang, J.G., Yau, W.Y., Suwandy, A., 2008b. Person recognition by fusing palmprint
(Gabor, 1946; Otsu, 1979; Vein recognition in Europe, 2004).
and palm vein images based on laplacianpalm feature. Pattern Recogn. 41 (5),
1514–1527.
References Sasaki, S., Kawai, H., Wakabayashi, A., 2005. Business Expansion of Palm Vein
Pattern. Authentication Technology. Fujitsu Sci. Tech. J. 41 (3), 341–347.
Chen, H., Lu, G., Wang, R., 2009. A new palm vein matching method based on ICP Sonka, M., Hlavac, V., Boyle, R., 1999. Image Processing, Analysis, and Machine
algorithm. In: Proceedings of the International Conference on Interaction Vision, 2nd ed. New York.
Sciences pp.1207–1211. Zhang, D., Kong, W.K., You, J., Wong, M., 2003. Online Palmprint Identification. IEEE
Crisan, S., Tarnovan, I.G., Crisan, T.E., 2010. Radiation optimization and image Trans. Pattern Anal. Mach. Intell. 25 (9), 1041–1050.
processing algorithms in the identification of hand vein patterns. Comput. Zhang, Y.B., Li, Q., You, J., Bhattacharya, P., 2007. Palm vein extraction and matching
Stand. Inter. 32 (3), 130–140. for personal authentication. Lect. Notes Comput. Sci. 4781, 154–164.
Daugman, J.G., 1980. Two-dimensional spectral analysis of cortical receptive field Zhou, Y., Kumar, A., 2011. Human Identification using Palm-Vein Images. IEEE Trans.
profiles. Vision. Res. 20, 847–856. Information Forensics and Security. In Press.
Daugman, J.G., 1993. High confidence visual recognition of persons by a test of
statistical independence. IEEE Trans. Pattern Anal. 15 (11), 1148–1161.