International Journal of Emerging Technology and Advanced Engineering
Website: www.ijetae.com (ISSN 2250-2459, ISO 9001:2008 Certified Journal, Volume 3, Issue 5, May 2013)
New Multimodal Biometric Approach with Two Score Level
Combination Strategies
Prateek Haksar1, Rama Gaur2
1
Pursuing Master of Technology, MEC, RTU, Kota
2
Associate Professor, MEC, RTU, Kota
Combination of Fingerprint and vein biometrics is an
attractive alternative in comparison to other biometrics.
[3] Vein pattern biometrics is unique feature to identify
individuals. The vein patterns could be obtained from
fingers or the whole palm. The equipments used to obtain
an image of the vain pattern are also simple and
inexpensive. Image processing techniques are capitalized
on to provide template creation and, at a later phase,
pattern matching. [4]
Abstract-- Biometrics is the science of verifying the
identity of an individual through physiological
measurements or behavioral traits. Since biometric
identifiers are associated permanently with the user they
are more reliable than token or knowledge based
authentication methods. The anatomy of human fingers is
quite complicated and largely responsible for the
individuality of fingerprints and finger veins. The
conventional level-1 fingerprint features, which illustrate
macro finger detail such as ridge flow and pattern type, can
be extracted from the low-resolution fingerprint images.
The acquired finger images are noisy with rotational and
translational variations resulting from unconstrained (pegfree) imaging. According to aforementioned issues my
paper focuses on a new approach for personal identification
that utilizes simultaneously acquired finger-vein and finger
surface (texture) images is presented. A new approach for
both the finger-vein and finger texture identification is
illustrated, which achieves significant improvement in the
performance over previously proposed approaches. The
finger-vein identification approach utilizes peg-free and
more user-friendly unconstrained imaging. I have used
Matlab R2012a for the simulation of finger and vein
recognition with the help of Image processing tools.
II.
The finger-vein identification approach utilizes pegfree and more user-friendly unconstrained imaging. The
proposed system simultaneously acquires the finger vein
and low resolution fingerprint images and combines these
two evidences using a novel score level combination
strategy. I examine the previously proposed finger vein
identification approaches and develop a new approach
that illustrates it superiority over prior published efforts.
The utility of low resolution fingerprint images acquired
from a webcam is examined to ascertain the matching
performance from such images. Miura et al. [4] have
further improved the performance for the vein
identification using repeated line tracking algorithm. The
robustness in the extraction of finger vein patterns can be
significantly improved with the usage of local maximum
curvature across the vein images and is detailed in
reference [5] with promising results.
Keywords-- Low resolution fingerprint, Finger vein
image, Gabor filter, Image enhancement, Score level
combination.
I.
MOTIVATION AND RELATED W ORK
INTRODUCTION
Digital image processing has a broad spectrum of
applications, such as remote sensing via satellites and
other spacecrafts, image transmission and storage for
business applications, medical processing, biometrics,
radar, sonar and acoustic image processing, robotics and
automated inspection of industrial parts. [1].
The use of biometrics within physical access control
(PAC) systems is one of the most broadly
commercialized sectors of biometrics, outside of forensic
applications. The requirements for the use of biometrics
within a larger physical access control system are
dependent on the interaction with existing access control
infrastructures. For this reason, the biometric system
must be designed to interface appropriately with a wide
range of access control systems [2]. Multimodal systems
employ combination of more than one biometric
recognition technique to arrive at a final decision. These
systems may be necessary to ensure accurate
performance.
III.
OUR W ORK
The Database is requested from Dr. Ajay Kumar, PQ
835, Department of Computing. The Hong Kong
Polytechnic University finger image database consists of
simultaneously acquired finger vein and finger surface
texture images from the male and female volunteers. This
database has been largely acquired during April 2009 March 2010 using a contactless imaging device in The
Hong Kong Polytechnic University campus. Our finger
vein identification approach utilizes peg-free and more
user friendly unconstrained imaging. The acquired
images are subjected to pre-processing steps that include
(a) Segmentation of region of interest (ROI)
(b) Translation and orientation alignment.
(c) Image enhancement to extract stable/reliable
vascular patterns.
575
International Journal of Emerging Technology and Advanced Engineering
Website: www.ijetae.com (ISSN 2250-2459, ISO 9001:2008 Certified Journal, Volume 3, Issue 5, May 2013)
Each of the acquired finger vein images is firstly
subjected to binarization, using a fixed threshold value as
230. In the complement of a binary image, zeros become
ones and ones become zeros; black and white are
reversed. Then the Sobel edge detector is applied to
entire image and the resulting edge map is subtracted
from the binarized image. Now, the isolated blobs are
removed eliminating number of connected white pixels
less than a threshold. [6]
For translation and orientation alignment an initial
point of the finger boundary from the left end is selected,
and then another boundary point on the same side is
sampled at a distance away from the starting point.
Hence the centroid is computed from using pair of these
points. The key purpose off finger boundary estimation is
to ensure that the rotational alignment of finger will be
carried out more precisely.
Finger vein feature extraction methods considered in
this work. The extraction of finger vein features using
repeated line tracking [4] and maximum curvature [5] has
been suggested with promising results. In this paper, we
systematically develop a new approach for the finger
vein feature extraction using Gabor filters.
We propose to formulate the framework for the finger
vein feature extraction using multi-orientation Gabor
filters. 2-D gabor filter is a linear filter whose impulse
response is designed by a harmonic function multiplied
by a Gaussian function. It uses the convolution property.
The venous patterns in the combined output image are
subjected to morphological operations to further enhance
the clarity of vein patterns. The top-hat operation filters
the grayscale or binary image. It computes the opening of
image and then subtracts the result from the original
image. The matching scores between two finger vein
feature vectors should be robust to accommodate the
translational and rotational variations in the normalized
vein images. These variations are often caused by
inaccurate (non-ideal) presentation of fingers in our
imaging setup or due to the inaccurate localization and
normalization. Therefore the matching score scheme
devised in this work attempts to compute the best
matching scores between two images while accounting
for such possible spatial shifts and rotation.
Now for finger texture image pre-processing images
acquired from the webcam (640 × 480 pixels) are firstly
automatically reduced to 580 × 380 pixels gray level
images since the cropped part does not provide any
useful finger details. This reduced size gray level image
is employed for the preprocessing. A Sobel edge detector
is firstly used to obtain the edge map and localize the
finger boundaries. This edge map also illustrates isolated
noise which is eliminated from the area thresholding.
Now, the image is enhanced by firstly subjecting the
image to median filtering to eliminate the impulsive
noise often present in the webcam acquired image. The
resulting images have low contrast and uneven
illumination. Therefore we obtain the background
illumination image from the average of pixels in 110×10
pixels image sub-blocks and bi-cubic interpolation. The
resulting image is subtracted from the median filtered
finger texture image and then subjected to histogram
equalization.
The enhanced knuckle image mainly consists of
curved lines and creases. Knuckle curved lines and
creases are to be detected. Knuckle features are then
extracted.[8] Algorithms are to be employed for the
matching of two knuckle features. Such localized
matching scores should be more effectively accounted
while matching low resolution finger texture images.
The estimated orientation ω of the image is used for
rotational alignment of the region of interest in vein
image. [7]
The normalized second order moment’s α12, α11 and
α22 are firstly computed as follows: [7]
Where I and (gx, gy) respectively represents the image
and the position of centroid in the image.
576
International Journal of Emerging Technology and Advanced Engineering
Website: www.ijetae.com (ISSN 2250-2459, ISO 9001:2008 Certified Journal, Volume 3, Issue 5, May 2013)
Figure 1: Block diagram for personal identification using simultaneous finger vein and finger texture imaging. [6]
IV.
score, equal error rate holistic, holistic fusion matching,
equal error rate non-linear and non-linear fusion
matching. All the statics are shown in following table:
RESULT
The database (10 images) obtained from The Hong
Kong Polytechnic University has given the following
result on basis of vein matching score, texture matching
1
Vein
Matching
Score
0.929962
2
0.928609
0.21
1.2396
98.760393
1.6839
98.316131
3
0.922782
0.29
1.2913
98.708666
1.8758
98.124222
4
0.869624
0.23
1.1568
98.843243
1.7154
98.284602
5
0.91.940
0.23
1.2249
98.775092
1.7267
98.273316
6
0.924962
0.22
1.2411
98.758925
1.7066
98.293374
7
0.906917
0.21
1.2028
98.79719
1.6782
98.321835
8
0.914699
0.31
1.2928
98.707242
1.9231
98.076946
9
0.925789
0.17
1.2039
98.796125
1.5896
98.410361
10
0.918684
0.25
1.2534
98.74655
1.7768
98.223161
Sr.
No.
Texture Matching
Score
EERH
Holistic Fusion
Matching
EERN
Non- Linear
Fusion Matching
0.17
1.211
98.789017
1.5907
98.409328
The statics here shown are taken by precision.
Resultant Images:
Step-by-step illustration of images we get during the
whole process. Firstly, finger vein after that fingerprint
images are shown.
There are total 64 images among them main images in
figure 2 (Vein) and figure 3 (Texture) is shown.
577
International Journal of Emerging Technology and Advanced Engineering
Website: www.ijetae.com (ISSN 2250-2459, ISO 9001:2008 Certified Journal, Volume 3, Issue 5, May 2013)
(a)
(b)
(d)
(g)
(c)
(e)
(f)
(h)
(i)
Figure 2: (a)Original Vein Image(b)Binarised Vein Image (c)Edge Map Subtracted Image (d)ROI mask (e)ROI Finger Vein Image (f)Enhanced
Finger Vein Image (g)Feature Extracted Vein Image (h)Vein To Be matched Up to 10 images (i) Vein Matching
(a)
(b)
(d)
(e)
(c)
(f)
(g)
Figure 3: (a) Edge Map (b) Image After Area Threshold (c) Rectangular Region (d)Enhanced Image (e)Feature Extraction (f)Texture to be
matched up to 10 images (g) Matching
578
International Journal of Emerging Technology and Advanced Engineering
Website: www.ijetae.com (ISSN 2250-2459, ISO 9001:2008 Certified Journal, Volume 3, Issue 5, May 2013)
[5]
REFERENCES
[1]
[2]
[3]
[4]
Fundamental of digital image processing, Anil K. Jain, Prentice
Hall, 1989.
Encyclopedia of Biometrics, Stan Z. Li (Ed.), Springer, 2009.
Hatim A. Aboalsamh, “A Multi Biometric System Using
Combined Vein and Fingerprint Identification”, International
Journal Of Circuits, Systems And Signal Processing
N. Miura, A. Nagasaka, and T. Miyatake, “Feature extraction of
finger-vein patterns based on repeated line tracking and its
application to personal identification,” Machine Vis.& Appl. pp.
194-203, Jul., 2004.
[6]
[7]
[8]
579
N. Miura, A. Nagasaka, and T. Miyatake, “Extraction of fingervein patterns using maximum curvature points in image profiles,”
Proc. IAPR Conf. Machine Vis. & Appl. pp. 347-350, Tsukuba
Science City, May 2005.
Ajay Kumar and Yingbo Zhou, “Human Identification using
Finger Images”, IEEE Trans. Image Processing, April 2012
A. Kumar and D. Zhang, “Personal recognition using hand-shape
and texture,” IEEE Trans. Image Process., vol. 15, no. 8, pp.
2454-2461, Aug. 2006
Shubhangi
Neware,
Dr.
Kamal
Mehta,
Dr.
A.S.
Zadgaonkar,“Finger Knuckle Surface Biometrics”, International
Journal of Emerging Technology and Advanced Engineering,
Volume 2, Issue 12, December 2012.