(IJCSIS) International Journal of Computer Science and Information Security,

Vol.6, No. 2, 2009

Making Palm Print Matching Mobile

Fang Li Maylor K.H. Leung Cheng Shao Chian
School of computer engineering School of computer engineering School of computer engineering
Nanyang Technological University Nanyang Technological University Nanyang Technological University
Singapore Singapore Singapore
[email protected] [email protected] [email protected]

Abstract— With the growing importance of personal print, voice, handwriting and DNA technologies joined in over
identification and authentication in today’s highly advanced the years.
world where most business and personal tasks are being replaced
by electronic means, the need for a technology that is able to Among the leading biometric technologies, fingerprint-
uniquely identify an individual and has high fraud resistance see based system is the most prominent and widely used biometric
the rise of biometric technologies. Making biometric-based technology, encompassing a market share of 58% in 2007 (A
solution mobile is a promising trend. A new RST invariant combine percentage of fingerprint and AFIS/Livescan
square-based palm print ROI extraction method was successfully technologies) [1]. The small size of the fingerprint-based
implemented and integrated into the current application suite. A device, ease of use and high accuracy has made it largely
new set of palm print image database captured using embedded
cameras in mobile phone was created to test its robustness. popular; however, as with most biometric solutions, it also has
Comparing to those extraction methods that are based on certain drawbacks. It is commonly found in most people that a
boundary tracking of the overall hand shape that has limitation layer of oil secretion or perspiration which emits from
of being unable to process palm print images that has one or microscopic pores residing on the tiny ridges of the fingers will
more fingers closed, the system can now effectively handle the cover the surface of the fingerprint areas. As the resolution
segmentation of palm print images with varying finger required for the fingerprint images are relatively high at
positioning. The high flexibility makes palm print matching approximately 500 dpi [7], this layer of secretion will render
mobile possible. the fingerprint image capturing device useless or less effective
in most cases. There are also cases whereby fingerprints wear
Keywords-Palm print, segmentation, mobility;
away due to work or fraudulently scarred, all these will lower
the effectiveness of fingerprint based systems.
I. INTRODUCTION
In this project, we explore a relatively new biometric
Personal identification and authentication have become a technology that employs palm print as the physiological
common task in today’s highly advanced world where more characteristic that is used to differentiate between each unique
and more day-to-day personal and business activities have been individual. Palm prints are rich in features such as principal
computerized [1-3]. Traditional identification and lines, wrinkles, ridge, datum points and minutiae points, all of
authentication systems rely on either a token item (e.g. a which could be extracted at relative low resolution. Palm prints
security pass card) or some knowledge only the user would also have a much larger surface area as compared to
know (e.g. passwords). Such systems are usually expensive in fingerprints, which indicates that more features could be
terms of time and resources to maintain and expand its usage. extracted from it, adding higher level of accuracy to it. These
The most critical flaw of these systems is that since they do not advantages place palm print-based technology as a promising
use any inherent characteristics or attributes of the individual biometric identification system.
user, they are unable to differentiate between an authorized
personnel and an impostor who have fraudulently come to Palm print recognition is an effective biometric technology
possess the token or knowledge (such as stolen credit card or that is gaining widespread acceptance and interest from
lost password). As such, these problems have led to system researchers all over the world. As with most other biometric
developers and researchers to explore into alternative solutions, technologies, the process of palm print identification includes
and thus the intensified research arises on biometric various stages from data acquisition, data pre-processing,
identification and authentication systems [4-6]. feature extraction to matching process.
Following this initial foray into biometric research, several The main aim of this research is to improve the (Region Of
forms of biometric systems based on different physiological or Interest) ROI extraction process to increase the system
behavioral characteristics have been developed. The first robustness. By implementing and integrating a new square-
commercial system, Identimat was developed in the 1970s [7]. based palm print ROI method into the previous application
The system was based on the measurement of the shape of the suite, the system is now able to overcome the limiting problem
hand and the lengths of the fingers as the basis for personal of failure to process palm print images with closed fingers, at
identification. After that, various forms of biometric systems the same time, to be (Rotation, Scaling and Translation) RST
such as fingerprint-based systems and iris, retina, face, palm invariant, thus increase the flexibility of the system and in turn

1 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
 (IJCSIS) International Journal of Computer Science and Information Security,
Vol.6, No. 2, 2009
open up the possibility of bringing the palm print technology digital image stored in the system. Such digital device can be a
mobile. A new set of palm print image database captured using digital camera [8], which is shown in Figure 2, or scanner [9],
embedded cameras in mobile phone was created to find most which is shown in Figure 3. This method is regarded as an
robust ROI extraction techniques. “on-line” method.

II. OVERVIEW OF PALM PRINT MATCHING SYSTEM

Palm print recognition is an effective biometric technology
that is gaining widespread acceptance and interest from
researchers all over the world. As with most other biometric
technologies, the process of palm print identification includes
various stages from data acquisition, data pre-processing,
feature extraction to matching process. The system overview
is shown in Figure 1.

Figure 2. Acquisition of a typical image sample using digital camera [8].

Figure 3. Acquisition of a typical image sample using scanner [9].

Inked palm print is not a good choice for mobile matching,

so we only discuss online palm print in this research.

IV. PRE-PROCESSING:REGION OF INTEREST (ROI)

EXTRACTION
After image acquisition, the raw input is passed to the
Figure 1. Palm print matching system overview. verification stages to perform various image processing
operations. Normally, the raw image consists of palm, fingers,
wrist, and a substantial amount of background area. For the
verification process, only the inner area of the palm is of
III. IMAGE ACQUISITION interest. The system needs to trim away those unwanted
In image acquisition stage, an image of the user’s palm is portions of the raw image to reduce the amount of computation
captured by the system. Palm image can be acquired by a few required in the subsequent stages. Another problem with the
methods. One method is to apply a uniform layer of ink on the raw input image is that the location and orientation of the palm
palm and place the palm on a paper; the paper is then scanned is not fixed.
into PC to obtain a digital image. This method is regarded as Palm prints may show certain degree of distortion as the
an “off-line” method. image may be captured at different times and rotated at
Another method for palm image acquisition is to use digital different angles. Furthermore, it could also be affected by
devices to photograph the palm and immediately obtain a varying conditions in terms of temperature, humidity and

2 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
 (IJCSIS) International Journal of Computer Science and Information Security,
Vol.6, No. 2, 2009
lighting condition. As such, even if two images are from the utilizes pegs [10-13] to restrict the movement and positioning
same palm, we could end up with a conclusion that they are of the hand during the acquisition process in order to improve
from different sources. the image quality, to solve RST problem, and to ensure that the
fingers are properly separated.
Palm print preprocessing, the segmentation process,
involving the correction of such distortion and placing all the
palm prints in the database under the same coordinate system
and orientation such that the proper expected area of each palm
print can be extracted for use in accurate feature extraction and
matching, greatly improves the efficiency and correctness of
the identification system.
The output of palm print segmentation is a sub image,
known as the Region of Interest (ROI) or central part sub-
image of the palm print, which is cut out from the original
input image. This sub image represents the inner area of the
palm, where most of the palm print features are within this
area. Those methods can be further classified into two different
classes: square-based ROI extraction and inscribed circle-based
ROI extraction [8]. As the circle-based approach consumes a
significantly higher amount of computation resources, based on
this experimental outset, we will only focus on square-based
ROI extraction approach throughout this research.
The basic idea of square-based ROI extraction technique as Figure 5. Palm image acquisition device with pegs [10].
demonstrated in Figure 4 is to determine key gaps-between-
fingers point on a palm print, thereafter two selected key points
are lined up to form the y-axis, subsequently, a second line,
which is the x-axis is drawn perpendicular to the y-axis through
the middle point to form the origin. Finally, a square with a
fixed size, ROI, is extracted under this coordinate system. All
the pixels within this ROI are retained for further processing
whereas the area outside the window is ignored and discarded.
The essential rule in this extraction process is that the portion
of the image extracted should be available in all palm prints
from the database and there are sufficient palm print features
for extraction and comparison. Moreover, the extraction should
be RST invariant and gesture difference tolerant.

Figure 6. Palm image acquisition device with pegs and fixed ROI location
[11].

PolyU-Online-Palm print-II is a benchmark palm print

database for research purpose [5]. In the database, all the palm
Figure 4. Basic idea of square-based segmentation technique. images are captured by a specially designed device [13] as
All ROI extraction techniques principally rely on the shown in Figure 7. The distance from palm to camera, pegs
determination of the key gaps-between-fingers to draw up the between fingers, and the board between middle and ring fingers
coordinate system through the use of boundary tracking are fixed. A lot of researchers use this benchmark database and
algorithm. Thus this imposes a limitation in that for a palm the device constraint ROI extraction process based on this
print image to be properly segmented, the fingers in the image database is shown in Figure 8 [12].
need to be sufficiently separated in order for the boundary
tracing to work, and thus the key points are determined
accurately.
Such requirement led to the development of image
acquisition devices, which are shown in Figure 5, 6, and 7, that

3 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
 (IJCSIS) International Journal of Computer Science and Information Security,
Vol.6, No. 2, 2009
In this project, some built in cameras in mobile phones are
used for image acquisition. The advantage of using those built
in cameras as input devices is that the user does not need to
purchase any device because of the popularity of mobile
phones. Nowadays, almost everyone has a mobile phone with
built in camera. Using these cameras, the palm print
verification system can be easily integrated into any security
systems without requiring extra image acquisition devices.

V. IMPROVED SQUARE-BASED ROI EXTRACTION METHOD

The aim of this research is to implement an improved ROI
extraction technique that is robust enough to overcome this
reliance of a standard image acquisition device, and thus able
to make use of the ubiquitous digital cameras and embedded
cameras in mobile phones to perform the image capturing
process, which in turn, will widen the scope of applications for
palm print-based systems.
The improved square-based ROI extraction technique
consists of the following steps:
Figure 7. Patent Palm image acquisition device designed by D. Zhang et al
[13].

Step 1: Gray image to binary image.

Step 2: Contour of hand generation

The above two step are standard ones in all palm print
preprocessing system, so details are omitted here.

Step3: Straight Lines Extraction

As mentioned in section 4, the locations of three feature

points need to be detected in order to set up a coordinate
system for palm print alignment. These key points lie on the
bottom of valleys between fingers. By observing the line
pattern of the boundary image, the bottom of valley is a short
curve joining the edges of adjacent fingers. The key points are
best represented as the mid-points of those short curves.

To locate the mid-point, one method is to first find the line

(Lm) that divides the inter-finger space into halves, the
intersecting point between Lm and the bottom curves of the
valleys is one of the desired key points. Usually, the edges of
two adjacent fingers form a V-shape. An angle can be
established by extending the V-shape edges until they
intersect. The line Lm can be found by calculating the bisector
of such angle. The method is illustrated in Figure 9.

Figure 8. ROI extraction process based on database PolyU-Online-Palm

print-II [12].

However, such devices are normally fixed to a site, and too

bulky to move around for usage. This lack of mobility
consequently results in the restrictive applications of palm print
authentication system [14-16].

4 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
 (IJCSIS) International Journal of Computer Science and Information Security,
Vol.6, No. 2, 2009
greatly decreased, which reduces the difficulty of identifying
the finger edges.
In order to get longest possible lines which have the high
chance to represent the location of the fingers, broken line
connection algorithm describe in [18] is adopted here to offset
the possible broken lines issue. The details of the boundary of
the palm are not necessary to analyze, so tiny straight lines are
excluded.
Parallel Lines Grouping
Finding the parallel pairs in this method means finding the
two lines which are the edge of the finger. To find the parallel
lines, firstly one line must be taken and then check every other
line whether it is the parallel partner of the first line or not.
After that, the next line must be taken as the first line, and
Figure 9. Method of Locating Key Point. check the other line to find its parallel partner. It must be done
until all of the lines have been checked whether they have
Generally, it is not easy to locate the V-shape edges parallel partner or not. The extracted parallel pairs extracted
directly. The problem can be solved by finding the parallel from the palm image in Figure 9 are shown in Figure 10.
edges of each finger first, then for every two adjacent fingers,
select an appropriate edge from each finger, and use the two
selected edges to form a V-shape pair. This approach is
feasible since detecting parallel line pairs are much easier than
detecting V-shape line pairs directly.
Based on the method above, the task of locating the three
key points in the boundary image can be divided into four
steps:
Step 3.1: Straight lines extraction: Find the straight lines in
the edge image, select the long ones which have the potential
to be edges of fingers.
Step 3.2: Parallel lines grouping: Group the extracted long
lines into parallel pairs, each pair represents two edges of a
finger.
Figure 10. The parallel pairs.
Step 3.3: V-Shape lines grouping: Reorder the parallel
pairs and group the lines into V-shape pairs, each pair
represents the edges of two adjacent fingers.
V-shaped Lines Grouping
Step 3.4: Key points location: Form an angle for each V-
shape pair and calculate the bisector of each angle. Find the V-shaped lines grouping means detecting the two line in
intersecting points between the boundaries of inter-finger between two fingers. Firstly, the parallel pairs obtained from
valleys and the calculated bisectors. The intersecting points the previous algorithm must be sorted from the leftmost to the
represent the desired key points. rightmost. Since the parallel pairs obtained from the previous
algorithm are stored using the 2-D array, it is easy to sort the
parallel lines and get the V-shaped line pairs. To get the V-
shaped pairs, it is basically shifting all the lines by one in to
Straight lines extraction
the right as below:
The edge pixels of the binary image are traversed and the
- Sort the parallel line pairs, so that the line pairs are
contour representation of the edge image is generated.
stored in left to right order.
Contour is a compact way to represent the shape of an image.
- For each parallel pair Pi in the sorted array, form a V-
The edge pixels are formed into separate groups, where each
shape pair with the right edge of Pi and the left edge of
group represents a connected curve. The Dynamic Two-Strip
Pi+1 (i = 0..I-2, where I is the total number of parallel
(DYN2S) algorithm [17] is employed to perform the curve
pairs).
fitting operation. If a curve has only small variation (i.e. it is
closed to a straight line), then this curve will be reduced to a The result of this phase can be seen in the Figure 11. The
single line segment which approximates to the original curve. same color lines identify the lines in the same pair:
After processing, the number of line segments in the image is

5 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
 (IJCSIS) International Journal of Computer Science and Information Security,
Vol.6, No. 2, 2009

Figure 11. The V-shaped pairs.

Key Points Location

The location of the key points can be calculated with
utilizing the V-shaped pairs obtained from the previous stage.
The algorithm can be explained as below:
a) Extend the lines of the V-shaped pair until they
intersect each other
b) Create the new center lines of an inter-finger space
which has the bisector of angle formed by the two
lines of the corresponding V-shape pair. Figure 12. Exceptional cases occurred usually when the inter-space
between the finger does not wide enough.
c) The key points for palm print alignment can be
located by calculating the intersecting points between
the bottom curves of valleys and the center lines of the Selecting Main Key points
corresponding inter-finger spaces.
When the number of key points identified is more than 3, it
However the above algorithm cannot work correctly in means that the key point between thumb and index finger is
some images due to the exceptional cases in the v-shape pairs. also detected. The system must exclude it since the key points
As demonstrated in Figure 12, the exceptional cases occurred needed to locate the ROI are only K1, K2, and K3. It is
usually when the inter-space between the finger does not wide necessary to select three desired key points out of the four.
enough. Different person has different finger shape, when Among the four points, every three adjacent points can form a
people close their figures tightly, the gap shape in-between triangle. There are two such triangles in total. This is
two figures can be different. It can be either a parallel V-shape illustrated in Figure 13.
line pair with the same angles’ value or an intersected V-shape
pair intersected above the expected key point.
To overcome this issue, additional checking is added to
check which category the pair belongs to. Once it is
confirmed, proper action will be taken to set the center line in
above step b) as follows:
- If it is a parallel line pair, the center line is set to be
the middle line of these two parallel lines.
- If it is an intersected line pair, the center line is set to
be the line which divides the intersected angle into
two halves.

Figure 13. Triangles Formed by Four Key Points.

It is noticed that the triangle containing K4 has a greater

height than the other triangle, because K4 is relatively far

6 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
 (IJCSIS) International Journal of Computer Science and Information Security,
Vol.6, No. 2, 2009
away from the other points. Hence, the desired key points can Figure 15 illustrates the above steps through the first
be determined by selecting the triangle with smaller height. example in Figure 14. The line structure is highlighted that it
is much easier for human vision to tell.

Step 4: Establish the ROI Window

After the desired key points (K1, K2, and K3) are located,
the coordinate system can be established in order to create the
ROI of the palm using the way described in section 4. The size
of ROI is dynamically determined by the distance between K1
and K3. It makes the ROI extraction scale invariant. The
distance between the camera and palm does not affect the
range of the region we extract. It gives the user maximum
freedom. A few samples of ROIs extracted from palms with
flexible positions are shown in Figure 14. Despite the scale
and gesture, the consistent regions are extracted.

Figure 15. Sample of line extraction.

VII. EXPERIMENT AND RESULTS

As the main aim of this research is to implement RST
invariant ROI extraction method that is capable of handling
palm print images that are varying in terms of fingers
positioning and distance to camera. A new mobile palm print
database with one thousand five hundreds photos is formed.
In this research, palm print images are captured using three
mobile embedded cameras with different resolutions from two
different mobile phones. During the image capturing process,
no fixed pegs were used to restrict the movement, rotation and
stretching of the hands. Each device is used to capture images
of both hands from thirty subjects. Figure 16, 17, and 18 show
the sample images captured by different mobile phones. With
each hand, five photos are taken for each of the five different
positioning of the hands in order to test the robustness of the
algorithm.
Figure 14. ROIs extracted with flexible palm positions.

VI. LINE EXTRACTION

Principal lines and course wrinkles are more obvious than
original ROI for human vision to judge the efficiency of ROI
extraction. Line information is highlighted by following steps
[19]:
i. Apply the averaging mask
ii. Applying the line detection masks Figure 16. Palm print captured using D810 VGA camera.

iii. Threshold the image

iv. Line thinning

7 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
 (IJCSIS) International Journal of Computer Science and Information Security,
Vol.6, No. 2, 2009

Figure 17. Palm print captured using D810 2MP camera.

Figure 20. The images used for experiment with various positioning of
fingers.

Figure 18. Palm print captured using SGH-i900 5MP camera.

The complete procedure to extract ROI in this research is

illustrated by one example shown in Figure 19.

Figure 21. The extracted ROIs of palm prints in Figure 20.

To test the robustness of the new algorithm, 15 palm

images from one person used in the experiment. There is no
constraint on palm position and orientation. The subject can put
her palm freely while taking the pictures. As for gesture, these
palm images can be classified into 3 classes according to the
abnormality’s level of both the palm’s shape and position. The
first class consists of 5 normal-shaped palm images with
optimal in-space distance between the fingers. The second
class consists of 5 normal-shaped palm images with various
positions and in-space distance between the fingers. The last
class consists of 5 abnormal-shaped palm images with various
position and in-space distance between the fingers.
Figure 19. The complete procedure to extract ROI in this research. Figure 20 shows the images used in the experiment. Images
in the same row represent the same class according to the

8 http://sites.google.com/site/ijcsis/
ISSN 1947-5500
 (IJCSIS) International Journal of Computer Science and Information Security,
Vol.6, No. 2, 2009
sequence of class. E.g., the images in the first row are classified [3] T.Savič, N. Pavešić, Personal recognition based on an image of the
in to first class, and so on. palmar surface of the hand. Pattern Recognition, Vol. 40, Issue 11 , P
3152-3163, November 2007.
The results of all images are expected to be approximately [4] W.X. Li, D. Zhang and Z.Q. Xu. Image alignment based on invariant
the same, since they are taken from the same person’s palm. features for palm print identification. Signal Processing: Image
Communication. 2003, Vol. 18,
Figure 21 shows the result of ROI extraction of the images [5] H. K. Polytechnic University. Palm print database, 2005. Biometric
in Figure 20. Observe from Figure 21, the new technique is Research Center Website. [Online] 2005.
approved to be successful as the ROIs are acceptably http://www4.comp.polyu.edu.hk/~biometrics/.
consistent. [6] W.X. Li, Authenticating Personal Identites Using Palm Print
Recognition. s.l. : Hong Kong Polytechnic University, 2004.
Take image pair 11 and 14 as example. The images’ scales, [7] D. Zhang, Palmprint Data. Palmprint Authentication. s.l. : Springer
locations, and orientations vary a lot, the extract two ROIs are Science + Business Media Inc., 2004.
acceptably consistent. It shows that the new technique can [8] A. Kumar, D. Wong, H. Shen, A. Jain, Personal Verification using
tolerate RST variance well. Palmprint and Hand Geometry Biometric. Lecture Notes in Computer
Science, vol.2688, 2003.
Among the three gesture categories, the ROIs in same row [9] C. Hana; H. Chengb, C. Linb, K. Fanb, Personal authentication using
are very similar to each other. ROIs in second class are more palm-print features. Pattern Recognition, Vol. 36, P371-381, 2003.
similar to those in first class than third class. Although the [10] A. Ross, A. Jain, Information fusion in biometrics Pattern Recognition
results in third class are not perfect, they are still satisfactory Letters vol. 24 , P 2115–2125, 2003.
since they are consistently representing the similar significant [11] C. Han, A hand-based personal authentication using a coarse-to-fine
palm print features as other two classes. strategy. Image and Vision Computing vol. 22, P909–918, 2004.
[12] D. Zhang, W. Kong, J. You, M. Wong, Online Palmprint Identification.
IEEE Tranactionson Pattern Analysis and Machine Intelligence, vol. 25,
no. 9, September 2003.
VIII. CONCLUSION [13] W. Kong, D. Zhang, W. Li, Palmprint feature extraction using 2-D
The improved square-based palm print ROI extraction Gabor filters, Pattern Recognition vol. 36, P 2339 – 2347, 2003.
method was successfully implemented and integrated into the [14] D. Zhang, V., N. Luo, A. Kumar, Robust palmprint verification using
2D and 3D features. Pattern Recognition, vol. 43, Issue 1, P 358-368,
current application suite. In comparison to other ROI extraction January 2010.
methods that are based on boundary tracking of the overall
[15] Human identification in information systems: Management challenges
hand shape with limitation of being unable to process palm and public policy issues. Clarke, R. 1994.
print images that have one or more fingers closed, the system [16] Zhenhua Guo, David Zhang, Lei Zhang, Wangmeng Zuo, Palmprint
can now effectively handle the segmentation of palm print verification using binary orientation co-occurrence vector, Pattern
images with varying finger positioning. It opens up the Recognition Letters, Volume 30, Issue 13, P1219-1227, October 2009.
possibility of bringing the palm print technology mobile. [17] M. K. Leung and Y. H. Yang, “Dynamic Two-strip Algorithm in Curve
Fitting”, Pattern Recognition, Vol.23, pp.69-79, 1990.
Through the experiments and findings in this research, it [18] X. Wu, K. Wang, A novel approach of palm-line extraction, Proceedings
was found that the images captured with latest model of the of the third international conference on image and graphics, 2004.
mobile camera at five mega pixels, the verification rate was [19] R. C. Gonzalez and R. E. Woods, Digital Image Processing (Second
close and comparable to those of the palm print images in the Edition), Prentice-Hall Inc., New Jersey, 2002.
research benchmark. Thus, with the certain continuous [20] J. Wang, W. Yau, A. Suwandy, E. Sung, Person recognition by fusing
advancement in mobile camera technologies, it would be in no palmprint and palm vein images based on “Laplacianpalm”
time that the usage of palm print authentication in mobile representation, Pattern Recognition, vol. 41, Issue 5, P1514-1527, May
2008.
systems be proved as viable and practical for widespread
applications.
However, researchers have proved that the identification
performance of most of the unimodal methods are not
satisfactory due to a variety of problems such as noise in data,
restricted degrees of freedom, non-universality, and low
accuracy. Fusing several modalities together might be able to
make system more robust [20]. In future, we will investigate a
feature to work together with palm print without compromising
the key advantage of our current system, that is, high degree of
freedom.

REFERENCES

[1] Biometric technologies - an introduction. Intelligence, Acuity Market.

2007, Biometric Technology Today, p. 9.
[2] A. Kong, D. Zhang and G. Lu. A Study of Identical Twins’ Palm prints.
s.l. : Springer-Verlag Berlin Heidelberg, 2005.

9 http://sites.google.com/site/ijcsis/
ISSN 1947-5500