APPLICATION OF IMAGE ENHANCEMENT AND SEGMENTATION TECHNIQUES ON VEIN PATTERN FOR BETTER IDENTIFICATION Thesis submitted in Partial Fulfillment for the award of Degree of Doctor of Philosophy in Computer Science & Engineering By S.SHARAVANAN FACULTY OF ENGINEERING AND TECHNOLOGY VINAYAKA MISSIONS UNIVERSITY (VINAYAKA MISSIONS RESEARCH FOUNDATION DEEMED UNIVERSITY) SALEM, TAMILNADU, INDIA JANUARY 2016
208
Embed
APPLICATION OF IMAGE ENHANCEMENT AND ......1.1 Biometrics and Palm Prints 1 1.2 Need for Palm Print Technology 8 1.3 Biometric s Based Palm Print Verification Proces 9 1.4 Operation
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
APPLICATION OF IMAGE ENHANCEMENT AND
SEGMENTATION TECHNIQUES ON VEIN PATTERN FOR
BETTER IDENTIFICATION
Thesis submitted in
Partial Fulfillment for the award of
Degree of Doctor of Philosophy
in Computer Science & Engineering
By
S.SHARAVANAN
FACULTY OF ENGINEERING AND TECHNOLOGY
VINAYAKA MISSIONS UNIVERSITY
(VINAYAKA MISSIONS RESEARCH FOUNDATION DEEMED UNIVERSITY)
SALEM, TAMILNADU, INDIA
JANUARY 2016
VINAYAKA MISSIONS UNIVERSITY
SALEM
DECLARATION
I, S. Sharavanan, declare that the thesis entitled
APPLICATION OF IMAGE ENHANCEMENT AND
SEGMENTATION TECHNIQUES ON VEIN PATTERN FOR
BETTER IDENTIFICATION submitted by me for the Degree of Doctor
of Philosophy is the record of work carried out by me during the period
from 2009 to 2016 under the guidance of Dr. A. Nagappan, and has not
formed the basisfor the award of any degree, diploma, associate-ship,
fellowship, or other titles in this University or any other University or
Institution of higher learning.
Place: Salem
Date: Signature of the Candidate
VINAYAKA MISSIONS UNIVERSITY
SALEM
CERTIFICATE BY THE GUIDE
I, Dr. A. Nagappan, certify that the thesis entitled
APPLICATION OF IMAGE ENHANCEMENT AND
SEGMENTATION TECHNIQUES ON VEIN PATTERN FOR
BETTER IDENTIFICATION submitted for the Degree of Doctor of
Philosophy by Mr.S.Sharavanan, is the record of research work carried out
by him during the period from 2009 to 2016 under my guidance and
supervision and that this work has not formed the basis for the award of
any degree, diploma, associate-ship, fellowship or other titles in
this University or any other University or Institution of higher learning.
Place: Salem
Date: Signature of the Supervisor
iii
ABSTRACT
Many techniques have been proposed by the researchers earlier for
the identification of palm vein for authentication purpose. A set of
methods for the identification and authentication of Palm Veins ( PVs)
which uses various components of the PV images have been proposed for
better accuracy of identification and extraction. Also these methods have
to reduce the identification time with large number of input palm images. It
uses various geometric, wavelet features for the extraction of features of
PV and for the classification of PV used support vector machines which
produces high accuracy in classification.
In recent times, biometrics such as PVs, finger prints, face and iris
recognition have been extensively used in many employments together with
entry admission management, human being authentication for computers,
online banking, ATM‘s and foreign Transaction managements. PV
identification uses the exclusive prototypes of PVs to recognize the persons
at a sky-scraping stage of accuracy. This thesis offers a novel algorithm for
PV identification.
iv
A multi variant volumetric measure to perform palm vein
recognition is proposed. The method normalizes the image by resizing the
image and applies wavelet transform to increase the signal levels. The
transformed image is used to generate number of integral image and for
each integral image a set of Junction points and their coordinates are
identified. The identified features are presented as PV matrix and using
them, the Junctional volume and special volume to compute the
trustworthy measure of the PV given are computed. This method produces
efficient results in the false acceptance rate by reducing it. Also it improves
the accuracy of palm vein identification and authentication. This method
reduces the overall time complexity which is higher in other approaches.
A multi-level dorsal-deep Vein Pattern (VP) based PV recognition
approach is proposed. The method removes the noise and performs
histogram equalization to enhance the image. The enhanced image is
applied with wavelet analysis and splits the higher order and lower order
VP. Generated two different images are split into sub sample images and
their junction points are identified. Identified junction point matrix is used
to compute the dorsal depth and deep vein depth measures to compute the
cumulative weight. Based on cumulative weight an average distance
v
measure is computed to identify the person based on some threshold
value. The proposed method has produced efficient results and reduces
the false ratio and time complexity.
The two proposed approaches for the development of PV
authentication technology have been tested with different number of
classes and samples and produced efficient results in all the factors of
quality of PV recognition and authentication. Experimental results shows
the comparison of PV authentication accuracy produced by different
methods and it shows the proposed methods have produced 96%
accuracy which is better than the other methods at different number of
classes and samples.
vi
ACKNOWLEDGEMENT
I thank God Almighty who has been showering his blessings on
me bestowed strength, knowledge and courage all these days.
I express my sincere gratitude to our Founder of
Vinayaka Missions University, Dr. A. Shanmugasundaram and
I am grateful to respectable Madam Founder Chairman
Mrs. Annapoorani Shanmugasundaram for constant support.
I convey my sincere gratitude to our Chancellor Dr. A. S.
Ganesan and Vice - Chairman Dato Sri‘ Dr. S. Sharavanan, for
permitting me to do this research at this great institution V.M.K.V
Engineering College.
I would like to convey my thanks and gratitude to my guide and
philosopher, Dr. A. Nagappan, Principal, V.M.K.V Engineering College,
Salem, for having guided me in every aspect to complete the research
and thesis. I learned a lot from him. His positive attitude energy and
ability always motivated me to perform. His advice which I always
remember is ―Learn from Experience and Improve Your Work‖.
vii
This advice is always helped me not only to this research but also in
other aspects of my life.
My special thanks to our Vice Presidents Mr. J. Sathish Kumar
and Mr. N. V. Chandrasekar, Vice Chancellor Dr. V. R. Rajendran,
Registrar Dr. Y. Abraham and Dean (Research) Dr. K. Rajendran, of
Vinayaka Missions University, Salem, and to my colleagues and well
wishers who have helped me in one way or other in doing this research.
I would like to thank for CIE Biometrics for providing PUT Vein
Database, without that analysis on vein pattern would not be possible.
Last but not the least, I thank my parents, my wife and children
who were supporting me day in and day out during the course of my
research.
(S.SHARAVANAN)
TABLE OF CONTENTS
CHAPTER
NO.
TITLE PAGE
NO.
ABSTRACT iii
LIST OF TABLES
xiii
LIST OF FIGURES
xiv
LIST OF ABBREVIATIONS
xvii
1
INTRODUCTION
1
1.1 Biometrics and Palm Prints
1
1.2 Need for Palm Print Technology
8
1.3 Biometrics Based Palm Print Verification Process
9
1.4 Operation Modes of Biometric System
11
1.5 Advantages of Palm Print Biometrics
12
1.6 Disadvantages of Palm print Biometrics
14
1.7 PV Patterns
14
1.8 PV Authentication Technology
15
1.9 Principles of Vascular Pattern Authentication
15
1.10 Applications of Biometric Systems
17
1.11 Authentication With PV Images
22
1.12 Details of Technology 23
1.13 PV Acquisition Methods
24
1.14 Organization of The Thesis
25
1.15 Objective
26
2
LITERATURE SURVEY
28
3
PALM VEIN RECOGNITION SYSTEM USING
LOCAL BINARY PATTERN AND GABOR
FILTER USING CLAHE BASED CONTRAST
ENHANCEMENT METHOD
54
3.1 Introduction
54
3.2 Previous Research
55
3.3 Block Diagram
57
3.4 Methodologies
57
3.5 Input, Selected Region and Results of
Combine
Features 58
3.6 Architecture of PV Recognition 75
3.7 Junction Point (JP) Detection 78
3.8 Cross Correlation on Join Point Extraction 78
3.9 Limiting The PV Region 79
3.10 Extending and Sub-Sampling The Contained
Region 80
3.11 Extracting The PV Code By Using the Local
Binary Pattern 81
3.12 Matching The Extracted Codes With Enrolled
ones 83
3.13 Extracting Wavelet Transformed Feature:
Global Features 84
3.14 SVM Classification 86
3.15 Experimental Results 87
3.16 Summary 88
4 MULTI-VARIANT VOLUMETRIC MEASURE
ON UPPER EXTREMITYVP BASED PV
RECOGNITION USING WAVELET
TRANSFORM 91
4.1 Introduction 91
4.2 Methods Explored 94
4.3 Overview of Multi-Variant Volumetric Approach 99
4.4 Normalization 100
4.5 Wavelet Transform on Input Image 101
4.6 Canny Edge Detection
102
4.7 Integral Image Generation
102
4.8 Feature Extraction
103
4.9 Junction Point Identification
103
4.10 Junctional Volume Computation
110
4.11 Algorithm of Junctional Volume Computation
110
4.12 Spacial Volume Computation
111
4.13 Trustworthy Measure Computation
113
4.14 Summary
114
5
MULTI LEVEL DORSAL-DEEP VP BASED
PV AUTHENTICATION USING WAVELET
TRANSFORM
115
5.1 Introduction
115
5.2 Overview of Dorsal-Deep VP Based Approach
120
5.3 Noise Removal
123
5.4 Histogram Equalization
124
5.5 Wavelet Analysis
125
5.6 Sub-Sampling Image Generation
127
5.7 Junction Point Computation 127
5.8 Dorsal Depth Measure
129
5.9 Deep Vein Depth Measure
131
5.10 PV Recognition
132
5.11 Summary
134
6
RESULTS AND DISCUSSION
135
6.1 Multi - Variant Volumetric Measure on Upper
Extremity VP Based PV Recognition Using
Wavelet Transform 136
6.2 Multi - Level Dorsal - Deep VP Based PV
authentication Using Wavelet Transform 152
6.3 Comparative Analysis 163
7 CONCLUSION AND FUTURE WORK 165
8 REFERENCES 169
9 LIST OF PUBLICATIONS 189
LIST OF TABLES
TABLE NO.
Table 2.1
TITLE
Analysis of Various Techniques
PAGE NO.
53
Table 3.1
The Accuracy Rate of PV Images
88
Table 4.1
Displays the Values of Junction Point Matrix
109
Table 6.1
Details of Data Set Being Used
135
Table 6.2
Comparison of Resilience, Rotation and Noise
152
xiii
xiv
LIST OF FIGURES
FIGURE NO.
Figure 1.1
TITLE
Block Diagram of Biometric Verification System
PAGE NO.
10
Figure 1.2
PV Patterns
16
Figure 1.3
ATM with PV Recognition System
18
Figure 1.4
ATM with Small PV Authentication System
19
Figure 1.5 PV across Control Unit 21
Figure 3.1
Block Diagram of Gabor Filter and Local Binary
Pattern
57
Figure 3.2
Examples of Input PV Region
58
Figure 3.3
Examples of Selected PV Region
59
Figure 3.4
VP
60
Figure 3.5
Curvelet Decomposition Pattern
61
Figure 3.6
Gabor Texture Representation Regions
62
Figure 3.7
LBP Texture Representations
63
Figure 3.8
LBP and Gabor Performance
65
Figure 3.9
Individual Samples of Recognition
67
Figure 3.10
Samples Vary with Recognition Percentage – Set 1
68
Figure 3.11
Samples Vary with Recognition Percentage – Set 2
69
Figure 3.12
Comparisons of PCA and Gabor
70
xv
Figure 3.13 Comparisons of PCA and LBP 71
Figure 3.14 Comparisons of PCA and Gabor, LBP 72
Figure 3.15
Comparisons of PCA and Gabor, LBP With
Version Number of Samples
73
Figure 3.16
Comparison of Other Methods with proposed
74
Figure 3.17 Architecture of PV Recognition 77
Figure 3.18 Examples of Localizing the PV Region with Masks 80
Figure 3.19 Stretched Images of Figure 81
Figure 3.20
The LBP Operator
82
Figure 3.21
Graphical Representation of Accuracy Performance
89
Figure 3.22
Graphical Representation of Processing Time
90
Figure 4.1
Palm VP of Hand
92
Figure 4.2
Abstract VP of Human Hand
93
Figure 4.3
Proposed System Architecture-I
100
Figure 4.4
Block Diagram of Normalization
101
Figure 5.1
Displays the Abstract VP
116
Figure 5.2
Proposed System Architecture-II
122
Figure 6.1
Snapshot of Input PV Image Selected
137
Figure 6.2
Snapshot of Boundary Marked
138
xvi
Figure 6.3 Rotated Snapshot of PV Image and the
Region Marked to be Extracted
139
Figure 6.4
Snapshot of Extracted Region of Interest
140
Figure 6.5 Snapshot of Noise Removed Image 141
Figure 6.6
Snapshot After Background Removal
142
Figure 6.7
Snapshot of Normalized PV Image
143
Figure 6.8
Snapshot of Skeleton Identified Image
144
Figure 6.9
Snapshot of Identified Junction Points in the Image
145
Figure 6.10
Snapshot of PV Image Matched
146
Figure 6.11
Snapshot of Step by Step Result of Proposed
147
Figure 6.12
Snapshot of Input Image Selected for PV
Recognition
153
Figure 6.13
Snapshot of Region Extracted
154
Figure 6.14
Snapshot of Histogram Equalized ROI Image
155
Figure 6.15
Snapshot of Background Subtraction
156
Figure 6.16
Snapshot of Normalized Image
157
Figure 6.17
Snapshot of Junction Point Identified Image
158
Figure 6.18
Snapshot of Identified PV Image
159
xvii
LIST OF ABBREVIATIONS
ANN Artificial Neural Network
ASIFT Affine-SIFT
ED Euclidean Distance
EER Equal Error Rate
HD Hamming Distance
JP Junction Point
LBP Local Binary Pattern
LDP Local Derivative Pattern
NBI Normalized Back scattered Intensity
NIR Near InfraRed
PCA Principal Component Analysis
PV Palm Vein
ROI Region of Interest
SIFT Scale Invariant Feature
SVM Support Vector Machine
SURF Speeded-Up Robust Features
VP Vein Pattern
1
CHAPTER 1
INTRODUCTION
1.1 Biometrics and Palm Prints
Today, in our daily life, we are often being asked for verification of
our identity. Normally, this is done through the use of passwords when
pursuing activities like domain accesses, single sign-on, application logon
etc. In the process, the role of personal identification and verification
becomes increasingly important in our society. With the onslaught of
improved forgery and identity methods of impersonation, correct
authentication in previous ways is not sufficient. Therefore, new ways of
efficiently proving the authenticity of an identity at a low cost are heavily
needed. Various ways of approach have been explored to provide a solution
and biometric-based identification is proved to be an accurate and efficient
answer to the problem. Biometrics has been an emerging area of research in
the recent years and is devoted to identification of individuals using physical
traits, such as few based on hand geometry, iris, face recognition, finger
prints, or voices. As unauthorized users are not able to display the same
unique physical properties to have a positive authentication, reliability will
be ensured. This is much better than the current methods of using
2
passwords, tokens or personal identification number (PINs). At the same
time it provides a cost effective convenient way of having nothing to carry
or remember.
Identity management becomes more sophisticated due to the
development of digital processing techniques. Whatever be the
authentication system there is a presence of digital verification process
exists and that may be using any of the human anatomical part like nose,
eyes, palms, etc. The palm print is one among them which is becoming
more popular now a days. There are many researches going on palm print
recognition for various requirements. The digital image of palm print shows
the internal structure of nerves in human palms which is unique for each
human and how it could be used to identify a person is the vital problem.
Computer-based personal identification, also can be said as biometrics
computing began in 1970s. At that time, ‗Identity‘, the first commercial
system was developed, which measured the shape of a hand and focused
particularly on finger length. In the meanwhile, finger print-based automatic
checking systems were widely used in enforcement of law. Retina based
systems and iris-based systems were introduced in the mid-1980s. Today's
speaker identification has its root in the technological achievements of the
3
1970s; while signature identification and facial recognition are relative
newcomers to the industry.
In the ubiquitous network society, Individuals can also easily access
their information anytime and anywhere can be obtained and people are
also faced with the risk that others can easily access the same
information anytime and anywhere. Due to this risk, personal identification
methodology, which can differentiate among registered users and
imposters, is of grave importance.
Currently, passwords, Personal Identification Numbers (4-digit PIN
numbers) or identification cards are used for personal identification.
However, there is every likelihood that cards can be stolen or forgotten,
guessing passwords and guessing numbers are possible. Biometric
authentication technology is used to solve these problems, which identifies
people based on their unique biological information and it deserves
attention. In biometric authentication, an account holder‘s behaviors or
body characteristics are registered in a database and then compared with
others who may try to access that account to see if the attempt is legitimate.
The term biometrics refers to a scientific discipline involving
automatic methods for recognizing (verifying or identifying) people based
4
on their physical and/or behavioral characteristics. Many biometric systems,
exploiting these methods to establish identity, have previously been
presented in the literature among them, methods which make use of
biometric characteristics such as finger prints, face, voice, iris, retina, hand
geometry, signature or palm prints are the most common. While a
considerable research effort is directed towards the development of fast,
robust, efficient and user-friendly biometric systems, some major problems
that are still need to be tackled before they can be deployed on a larger
scale. One of the major challenges, which is yet to be solved, includes
increasing the performance of biometric systems in terms of recognition.
Towards this goal, a recent trend has emerged such as the employment of
multi-modal biometric systems which establish identity either by
considering several biometric modalities (e.g., the face, the iris, palm
prints, voice etc.) or by combining the recognition results of several
algorithms performed on the same biometric sample. The valid solution
from such approaches for the problem of recognition performance,
unfortunately user-convenience gets decreased, as it requires a much greater
effort from the user to operate the system or it increases the time needed to
process a single user. Hence the remedy is worse. From this point of view,
other solutions capable of increasing the recognition rates and not
5
influencing the convenience of using the biometric systems should be found
out. One possibility of increasing the recognition performance is to closely
examine the feature normalization techniques, which form the criteria of
decreasing the error rates of biometric systems, but have so far been
largely omitted in most research papers on the subject of biometrics.
Generally, only a sentence or two is devoted to the employed normalization
technique, even though representation of feature normalization is crucial
step in the design of a biometric system. Always feature normalization
techniques have a great impact on the procedure of constructing user
templates (or models), i.e., mathematical representations of the feature
vectors extracted from several measurements of the biometric
characteristic (e.g., palm prints) acquired during the enrollment stage, and
consequently on how user-specific biometric characteristics are modeled.
They represent a faster and more efficient way of boosting the
recognition performance of biometric systems which does not significantly
increase the processing time of a user.
These days many applications of biometrics are being used or
considered worldwide. Most of the applications are still at the early stages
of testing process, and end users find it as an optional. Any circumstance
6
that allows an interaction between man and machine is capable of
incorporating biometrics. Such situations may fall into a specific range
of application areas such as computer desktops, laptops, wired & wireless
networks, online banking and immigration, enforcement of law,
telecommunication synchronous and asynchronous networks. Fraud is
an ever-increasing problem and security is becoming a must in many walks
of life. Though research on the issues of finger print identification and
speech recognition have drawn considerable attention over the last 25
years and recently issues on face recognition and iris-based verification
have been studied extensively there are still some limitations to the
existing applications. Some finger prints of few peoples get worn away
due to the hand-work and some are born with unclear finger prints. The
existing iris-based identification system has not been proved to be
adaptive to eastern people who have quite different iris patterns from
those of people from western. voice based identification systems and
Face identification systems are less accurate and easy to be imitated. Efforts
on improving the present personal identification methods are to continue
and meanwhile newmethods are under investigation.
7
The palm print uses the similar set of characteristics used for finger
print recognition. Characteristics like ridge flow, ridge characteristics and
ridge structure of the raised portion of the epidermis are adapted. The data
represented by these friction ridge impressions either originated from the
same source or could not be made by the same source.
Palm print is based on ridges, principal lines and wrinkles on the
surface of the palm. A palm print refers to an image acquired of the palm
region of the hand. It can be either an online image (i.e. taken by a
scanner, or CCD) or offline image where the image is taken with ink and
paper.
The palm itself consists of principal lines, wrinkles (secondary lines)
and epidermal ridges. It varies from a finger print in that it also contains
other information such as indents, texture and identification marks which
can be used when comparing one palm to another.
Palm prints can be used for scientific tests or techniques used in
connection with the detection of crime or commercial applications. Palm
prints are normally found at crime scenes as the result of the offender's
gloves slipping during the time of crime, and thus exposing part of the
unprotected hand.
8
1.2 Need for Palm Print Technology
Biometrics has been an emerging field of research in the recent years
and is devoted to identification of individuals using physical traits, such as
those based on iris or retinal scanniridgesng, face recognition, finger
prints, or voices. As unauthorized users are not able to display the same
unique physical properties to have a positive authentication, reliability will
be ensured. Palm print is preferred compared to other methods such as
finger print or iris because it is always identical and can be easily
captured using low resolution devices as well as contains additional
features such as principal lines. Iris input devices are expensive and the
method is intrusive as people might fear of adverse effects on their
eyes. Finger print identification requires high resolution capturing
devices and may not be suitable for all as some may be finger
deficient. Palm print is therefore suitable for everyone and it is also non-
intrusive as it does not require any personal information of the user. Palm
print images are captured by acquisition module and are fed into recognition
module for authentication.
Compared with face recognition palm prints are hardly affected by
age and accessories.
9
Compared with finger print identification images of palm print
contain more information and need only low resolution image capturing
devices which reduce the cost of the system.
Compared with iris recognition the palm print images can be
captured without intrusiveness as people might fear of adverse effects on
their eyes and cost effective. Hence it has become such an important and
rapidly developing biometrics technology over the last ten years. Limited
work has been submitted on palm print identification and validation, without
being affected by the importance of palm print features. The functions of
the system is done by projecting palm print images onto a feature space
that spans the significant variations among known images.
1.3 Biometric Based Palm Print Verification Process
Biometric system is basically a pattern recognition system which
identifies a person using psychological or emphasizing behavioral metrics.
The characteristics such as 3D hand geometry, finger print and palm print
are read into system using some scanners and sensors and return a result.
Any kind of biometric system (Figure 1.1) has four stages named
1. Data Acquisition
2. Preprocessing
10
3. Feature Extraction
4. Feature matching
Figure 1.1 Block Diagram of Biometric Verification System
Data acquisition
The first stage of biometric verification process where the input
signals or images are gathered using input devices such as scanners. The
quality of signal given as input or image plays a vital role because the
quality of result depends on the quality of input signals or images.
Preprocessing
At this stage the signal quality and image is improved using various
11
preprocessing stages like filtering, normalization, rotation, segmentation and
noise removal.
Filtering: This is the process of selection of pixels from set of pixels
and the selection of signal or pixel depends on the value of pixel or signal.
Noise removal: This is a procedure of avoiding incomplete signals
and pixels from further stage of processing.
Feature Extraction: The process of extracting stable properties of
intra- class difference and high intra-class difference. They are used to build
the template of the data base.
Feature matching: Is a matching procedure to compute matching
score with the featured template and master template.
1.4 Operation Modes of Biometric System
Any kind of biometric system has three operating modes:
Enrollment
This combines the first three stages of the biometric verification
systemnamely (data acquisition / data developing skills, preprocessing, and
feature extraction). Any user has to be enrolled before verification
into the system by data acquisition and the features have to be extracted
12
and then stored into the system.
Identification
This tells the matching process of the biometric system. It works to
find out the user with the biometric features obtained from the user and
match with other biometric templates available in the system. It initiates
identification process without knowing the identity of the user.
Verification
The verification process is done when an identification process goes
successfully and then there is searching the record to identify the person
about name, ID card and other attributes.
1.5 Advantages of Palm Print Biometrics
Since the palm area is much larger so that more distinctive features
can be captured and compared to finger prints. This makes it much more
suitable in the process of identification systems than finger prints.
Advantages of using the palm
In addition to the palm, vein authentication can be done using the
vascular pattern on the back of the hand or a finger. However, the pattern in
the PV is most complex and covers the widest area. Due to the palm has
13
no hair, it is easier to photograph its vascular pattern. The palm also has no
sufficient variations in skin color compared with fingers or the back of the
hand, where the color can darken in specific areas.
There are two methods of photographing veins: reflection and
transmission. Fujitsu implements the reflection method. The reflection
process illuminates the palm and photographs the light that is reflected back
from the palm, while photographs light of the transmission method passes
straight through the hand. Both methods capture the nearby infrared light
given off by the region used for identification after diffusion through the
hand. Such an important difference between the reflection method and
transmission method is how they respond to changes in the hand‘s light
transmittance. When the body gets cool due to a low ambient
temperature, the blood vessels in particular capillaries decreasing the flow
of blood throughout the body. This suits up the hand‘s light transmittance,
so light passes through it in much more easier way. If the transmittance is on
the higher end, the hand can become organic molecule with light and light
can easily pass through the hand. In the transmission process, this yields
results in a lighter, less-contrasted image in which vessels are difficult to
see. However, a higher level of light transmittance does not significantly
14
affect the level or contrast of the reflected light. Therefore as a result,
with the reflection method, the vessels can much more easily be seen even
when the hand/body is cool.
1.6 Disadvantages of Palm print Biometrics
The palm print scanners are generally bigger in size and expensive
since they need to capture a larger area than the finger prints scanners.
1.7 PV Patterns
Blood veins are formed during the first eight weeks of gestation in a
chaotic manner, influenced by the environment like mother‘s womb. This is
why VP is identical to each individual, even twins. Vein growth is
associated with a person‘s skeleton, and while capillary arrangement
continue to grow and change, vascular patterns are formed during birth
and do not change over the course of one‘s lifetime.
To scan the veins, an individual‘s hand is placed on the hand guide
(the plastic casing of the scanner device) and the VP is captured by lighting
the hand with near-infrared light. Veins consist of deoxidized
hemoglobin, an iron-containing coloring matter (pigment) in the blood
that carries oxygen throughout the body. These pigments absorb the near IR
light and reduce the reflection rate causing the veins to appear as a black
15
pattern. An every individual‘s scanned PV data (biometric template) is
encrypted for a protection and registered along with the other details in
his/her profile as a reference for future comparison.
1.8 PV authentication Technology
PV authentication is performed according to the comparison
performed between various patterns of human PV. The PVs are the lines
appear in the palm image with the blue lines and such patterns are extracted
and stored in the PV data base. Vascular patterns are generis to each and
every individual, according to Fujitsu research — even identical twins have
different patterns. And since the vascular patterns on the body exist inside,
they cannot be duplicated by means of photography, voice recording or
pattern of finger prints, thereby making this procedure of biometric
authentication more secure than others.
1.9 Principles of Vascular Pattern Authentication
Hemoglobin in the blood is oxygenated in the lungs and carries
oxygen to the tissues of the body through the arteries. After it gets release
its oxygen to the tissues, the deoxidized hemoglobin backs to the heart
through the veins. These two kinds of hemoglobin have distant rates of
absorbency.
16
Deoxidized hemoglobin absorbs light at a wavelength of about 760
nm in the near-infrared region. When the palm gets illuminated with near IR
light, unlikely images can be seen by the human‘s eye, the deoxidized
hemoglobin in the PVs absorb light, hence reflection rate gets reduced and
causing the veins to appear as a black pattern, based on this principle the
region used for authentication on vein is photographed with near-infrared.
Using image processing [Figure 1.2], light and the VP is extracted
and registered. The VP of the person being authenticated is then verified
against the preregistered pattern.
Figure 1.2 PV Patterns
17
1.10 Applications of Biometric Systems
In spite of product development for financial solutions financial
damage is caused by fraudulent withdrawals of money using identity
spoofing with fake bankcards has been fast increasing in last years, and this
has emerged as a significant problem in society. Therefore, rapid increase in
the number of lawsuits filed by victims of identity theft against financial
institutions for their failure to control information to be used only for
personal identification. ―Protection of Personal Information legal Act‖ force
into effect in Japan on 1st May 2005.
Financial institutions have been focusing on biometric authentication
together with IC (smart) cards as a way to reinforce the security of personal
identification. Vein authentication always providing two kinds of systems
for financial solutions, depending on the registered VPs are stored. In one
method, the VPs are stored on the server of a client-server system. The
benefits of this system are that it provides an integrated capability for
managing VPs and comparison processing. In the other type, a user‘s VP is
stored on an integrated circuit card, which is beneficial because users can
control access to their own VP. Suruga Bank uses the server type for their
18
financial solutions, and Tokyo-Mitsubishi is the bank uses the integrated
circuit card system.
Figure 1.3 ATM with PV Recognition System
In July 2004, to ensure customer security, Suruga Bank introduced its
―Bio-Security Deposit‖ — the world‘s first financial service to use Palm
Secure (Figure 1.3). These kinds of services provide high security for
customers using vein authentication, does not require a verification proof
like bankcard or passbook which are used to prevent withdrawals from
branches other than the registered branch and ATMs, hence as a result
minimizing the risk of fraudulent withdrawals. To open a Deposit
account with Bio-Security features, customers go to a bank and have their
PVs photographed at the counter. In order to make sure about the secure
data management, the PV data is stored only on the vein database server at
the branch office where the account is opened.
19
Figure 1.4 ATM with Small PV authentication System
In October 2004, The Bank of Tokyo-Mitsubishi4 launched its
―Super-Integrated Circuit Card Tokyo Mitsubishi VISA.‖ These types of
cards always have the functions of a bankcard, credit card, electronic money
and PV authentication. From a technical prospect and user-friendly point of
view, Tokyo - Mitsubishi Bank narrowed the biometric authentication
methods suitable for financial transactions to PVs, finger veins and finger
prints. The bank then mailed a feedback form to 1,000 customers
and surveyed an additional 1,000 (Thousand) customers who used the
devices in their branches. At the final stage, the bank decided to employ
Palm Secure because the technology was supported by the largest number
of people in the questionnaire. The Super-Integrated Circuit Card contains
the customer‘s PV data and vein authentication algorithms combines
and performs vein authentication by itself. This system is beneficial
because the customer‘s information is not stored at the bank.When a
20
customer applies for a Super-Integrated Circuit card, the banker sends
the card to address of the customer‘s home. To activate the PV
authentication function, the customer brings the verification card and his or
her passbook and seal to the bank counter, where the customer‘s
required vein information is registered on the card. After registration
process, the customer can make transactions at that branch‘s counter
and any ATM (Figure 1.4) using PV authentication and a matching PIN
number.
The Hiroshima Bank started this type of service in date of April 2005,
followed by The Bank of Ikeda6 in date of June 2005. Other
financial institutions, including The Nanto Bank, planned and organized
to start similar services during fiscal 2005.
In 2006, Fujitsu reduced the Palm Secure sensor to 1/4 of its current
size for its next generation product. By using a sensor on existing ATMs,
there will be room or place on the operating panel for a sensor for Felicia
mobiles, a 10-key pad that meets the Data Encryption Standard (DES), as
well as other devices including electronic calculator. The downsized sensor
can also be installed on ATMs in convenience stores.
21
In addition to product development for financial solutions, Fujitsu has
initiated to develop product applications for the general market. Two
products moving faster are in great demand on the general market place.
One product is for a physical access control unit that uses Palm Secure to
protect entrances and exits, and the other product is nothing but a logical
access control unit that uses Palm Secure to protect input and output of
electronic data.
Palm Secure units are used to control access to places containing
systems or machines that manage personal or other confidential and more
secured information, such as machine rooms in companies and outsourcing
centers where important customer data is kept.
Figure 1.5 PV across Control Unit
Due to increasing concerns about security, some commercial sectors
and homes have started using this system to enhance security and safety in a
day to day life. Considering both of these applications, the combined
22
form of the following advantages provides the optimum system: a hygienic
and contactless unit ideal for use in public places, simple user-friendly
operation that requires the user to simply hold a palm over the sensor, and
an authentication mechanism that provides impersonation difficult. The PV
authentication login unit controls access to electronically stored information
(Figure 1.5). When considering the units for financial solutions, there are
two types as follows: a server type and an Integrated Circuit card type.
Becausethe Palm Secure login unit can also be used for authentication
using conventional IDs and passwords, existing operating systems and
available applications can continue to be used. It is also possible to
develop the unit into an existing application to enhance operability. In
the initial stage of introduction, the units are having limitations like
areas of businesses handling personal information that came under the
―Protection of Personal Information Act‖ enforced in the date of April 2005.
However, usage of the units is now expanding to leading-edge businesses
that handle confidential information.
1.11 Authentication with PV Images
Unlike the previous vein feature based authentication mechanism, the
vein code based authentication mechanism extracts the feature codes from
23
palm vein imagery which is represented in binary way. The binarized data
set improves the efficiency of authentication as well as increases the speed
of authentication. From the single piece of PV image there are number of
feature codes can be generated and converted in to binary form. The
generated binary form data can be applied to various problems and can
provide uninterrupted service to many areas. The modern technology
extends the application and scope of PV authentication mechanism which
enables the possibility of using biometric authentication mechanism in a
dense organization.
The application of PV feature based authentication mechanism is
growing in day by day which uses biometric features and the feature is very
much unique for any person. This improves the essential secure storage of
bio features which can be shared between many application.
1.12 Details of Technology
PV image normalization technique
The PV image has to be normalized before being used to perform PV
authentication. The normalization method must be more efficient so that the
features of PV could be maintained. To perform such efficient
normalization, the Fujitsu Laboratories has developed an efficient method
24
which uses contour information to maintain the sh ape and position in PV
images. The image captured by the device attached is verified for the
position and shape and the method removes the distortion from the image
obtained.
Feature code extraction technology
The method uses a different size of feature code which is in the size of
2048 bits, a 2 byte style. The method first generates a sectional image which
fixed size and splits the image into number of regions. For each region the
method extract the features and according to the amount of information
present in the regional image, a 2 byte information or code is framed. The
generation of 2 byte information is done by compressing the information
present in the region. The region based approach enables the
identification of micro changes in the VPs which can be introduced at
different shapes and positions.
1.13 PV Acquisition Methods
There are many ways to snapshoot the PV but each differs with the
accuracy of the vein image.
25
High-Speed Image-Capture Technology for PV Biometric
authentication
Fujitsu Lab developed a prototype authentication device that
employs a high-speed shutter to capture images of the PVs without
blurring even when the palm is in motion, in contrast to the previous
version which captured images when the palm was suspended above the
sensor.
An improved PV identification based on thermal PV image
The infrared PV image is captured using the infrared waves and the
image is stored for processing.
1.14 Organization of the Thesis
This thesis is organized into seven chapters. Chapter 1 gives the
introduction to thesis with Biometrics concepts and palm prints for the
research work. Chapter 2 describes the literature survey related to
palm prints, Algorithms and Methods. Chapter 3 deals with the PV
recognition system using local binary pattern and Gabor filter using
Clahe based contrast enhancement method. In Chapter 4, the work is
Multi – Variant Volumetric Measure on Upper Extremity VP Based PV
Recognition Using Wavelet Transform.
26
Chapter 5 focuses on the Multi-Level Dorsal-Deep VP Based
PV authentication Using Wavelet Transform while Chapter 6 is
devoted to Results and Discussion and Chapter 7 forms the conclusion and
future work envisages respectively.
1.15 Objectives
Biometrics such as PVs, palm prints, face recognition and iris
identification have been extensively used in a lot of employments together
with entry admission management, human being authentication for
computers, online banking, ATMs and foreign Transaction
managements. PV identification uses the exclusive prototypes of PVs to
recognize the persons at a sky-scraping stage of precision. This
work offers two approaches for the development of PV authentication
technology namely A multi variant volumetric measure to perform PV
recognition and A multi- level dorsal-deep VP based PV recognition
approach Both the methods show very high accuracy and also less
processing time.
1. To study and apply appropriate image segmentation technique on VP.
2. To measure VP based PV recognition using wavelet transform.
27
3. To identify features present in PV matrix which is used to compute the
Junctional volume and special volume to find the trustworthy measure of
the PV given.
4. To achieve a better accuracy and low False Acceptance Rate (FAR).
5. To achieve less Processing Time compared to existing methods.
28
CHAPTER 2
LITERATURE SURVEY
The extended research of palm print has been done in many years and
there has been various methodologies have been proposed for the
identification of palm print of a particular person in this era. Some of the
unique methodologies here and their effectiveness and default with many
characteristics are explored here.
Junichi Hashimoto, 2006, discussed VPs based biometric
authentication approach for biometric authentication. Secured Smart Card
Using PV Biometric On-Card-Process [1], discuss the PV biometric
system and its compatibility in financial sector, software design for on-card-
processing solution based on Java virtual machine.
In both the paper the solution is stimulated and the result obtained
and that was tested on PC login application. This increases the tampering of
forgery and increases the quality of authentication. The security level of PV
Biometric On-Card-Process [1], is highly reliable since the FRR and FAR
is very low compared to other biometric systems.
A PCA based PV authentication system is presented in [3], which
uses the Princple component analysis method to perform feature extraction
29
and Yuhang Ding, Dayan Zhuang and Kejun Wang, July 2005[4], analyzed
the difficulty of hand vein recognition and propose a thresol segmentation
with thinning approach.
To capture the palm vein image the infrared camera is used and by
using the PCA approach the extracted features are converted into feature
vector. Both the papers using the highest information of varying size, the
pattern matching is performed to find out the best match from the data base
to perform authentication. The method extracts the edge and junction points
and then performs pattern matching to compute the distance. Based on
computed distance the method performs biometric authentication.
In combination of [3] and [4], there exist Shi Zhao, Yiding Wang
and Yunhong Wang, proposed [5] , uses a hand dorsa to extract the edge
features of palm vein to perform biometric authentication. The method
replaces the necessary of using high quality devices and reduces the cost of
biometric authentication.
In combination with above study Yi-Bo Zhang et al [16] , discusses a
palm vein authentication mechanism which also captures the palm vein
image through the infrared palm image capturing device and then it
identifies the region of interest. Once the region is being identified then the
method extracts the palm vein pattern which is obtained by applying multi
30
level filtering technique. Finally the extracted pattern is performed with
patten matching to perform biometric authentication.
Masaki Watanabe, Toshio Endoh, MoritoShiohara, and Shigeru [6]
and Shani Sarkar et al [12], performs a detailed inspection on the palm vein
authentication mechanisms presented earlier which uses the vessel pattern
of person hands. The growth of biometric authentication has great influence
on various sectors person identification in banks, markets and more.
Paper [6] have shown a biometric authentication using contactless PV
authentication device that uses blood vessel patterns as a personal
identifying factor. Implementation of these contactless identification
systems enables applications in public places or in environments where
hygiene standards are required, such as in medical applications. In addition,
sufficient consideration was given to individuals who are reluctant to come