04-10-2014, 11:20 AM
A DCT-based Local Feature Extraction
Algorithm for Palm-print Recognit on
[attachment=68820]
Abstract-
In this paper, a spectral feature extraction algorithm is proposed for palm-print recognit on, which can ef iciently capture the detail spatial variations in a palm-print image. The entire image is segmented into several spatial modules and the task of feature extraction is car ied out using two
dimensional discrete cosine transform (2D-DCT) within those spatial modules. A dominant spectral feature selection algorithm is proposed, which of ers an advantage of very low feature dimension and results in a very high within-clas compactnes and betwe n-clas separabil ty of the extracted
features. A principal component analysis is performed to further reduce the feature dimension. From our extensive experimentations on dif erent palm- print databases, it is found that the performance of the proposed method in terms of recognit on ac uracy and computational complexity is superior to
that of some of the recent methods.
1 INTRODUCTION
Conventional ID card and pas word based identif cation
methods, although very popular, are no more reliable as
before because of the use of several advanced techniques
of forgery and pas word-hacking. As an alternative, biometrics, such as palm-print, finger-print, face and ir s
being used for authentication and criminal identif cation [7]. The main advantage of biometrics is that these are not
prone to theft and los , and do not rely on the memory of
their users. Moreover, they do not change signif cantly over
time and it is dif icult for a person to alter own physiological
biometric or imitate that of another person's. Among
dif erent biometrics, in security ap lications with a scope of
col ecting digital identi y, the palm-prints are recently get ing
more at ention among researchers [3], [9]. Palm-print
recognit on is a complicated visual task even for humans. The primary dif iculty arises from the fact that dif erent
palm-print images of a particular person may vary largely, while those of dif erent persons may not neces arily vary
signif cantly. Moreover, some aspects of palm-prints, such
as variations in il umination, posit on, and scale, make the
recognit on task more complicated [6]. Palm-print
recognit on methods are based on extracting unique major
and minor line structures that remain stable throughout the
lifetime.
PROPOSED METHOD
For any type of biometric recognit on, the most important
task is to extract distinguishing features from
data, which directly dictates the recognit on ac uracy. In
comparison to person recognit on based on face or voice
biometrics, palm-print recognit on is very chal enging even
for a human being. For the case of palm-print recognit on, obtaining a signif cant feature space with respect to the
spatial variation in a palm-print image is very crucial. Moreover, a direct subjective cor espondence betwe n
palm-print features in the spatial domain and those in the
frequency domain is not very ap arent. In what fol ows, we
are going to demonstrate the proposed feature extraction
algorithm for palm-print recognit on, where spatial domain
local variation is extracted from frequency domain
transform.
5 CONCLUSIONS
In the proposed palm-print recognit on scheme, instead of
operating on the entire palm-print image at a time, dominant
DCT-based features are extracted separately from each of
the modules obtained by image-segmentation. It has be n
shown that because of modularization of the palm-print
image, the proposed dominant spectral features, that are
extracted from the sub-images, at ain bet er discriminating
capabil ties. The proposed feature extraction scheme is
shown to of er two-fold advantages. First, it can precisely
capture local variations that exist in the major and minor
lines of palm-print images, which plays an important role in
discriminating dif erent persons. Second, it util zes a very
low dimensional feature space for the recognit on task, which ensures lower computational burden. For the task of
clas if cation, an Euclidean distance based clas if er has
be n employed and it is found that, because of the quality
of the extracted features, such a simple clas if er can
provide a very satisfactory recognit on performance and
there is no ne d to employ any complicated clas if er. From
our extensive simulations on dif erent standard palm-print
databases, it has be n observed that he proposed method,
in comparison to some of the recent methods, provides
excel ent recognit on performance.
Algorithm for Palm-print Recognit on
[attachment=68820]
Abstract-
In this paper, a spectral feature extraction algorithm is proposed for palm-print recognit on, which can ef iciently capture the detail spatial variations in a palm-print image. The entire image is segmented into several spatial modules and the task of feature extraction is car ied out using two
dimensional discrete cosine transform (2D-DCT) within those spatial modules. A dominant spectral feature selection algorithm is proposed, which of ers an advantage of very low feature dimension and results in a very high within-clas compactnes and betwe n-clas separabil ty of the extracted
features. A principal component analysis is performed to further reduce the feature dimension. From our extensive experimentations on dif erent palm- print databases, it is found that the performance of the proposed method in terms of recognit on ac uracy and computational complexity is superior to
that of some of the recent methods.
1 INTRODUCTION
Conventional ID card and pas word based identif cation
methods, although very popular, are no more reliable as
before because of the use of several advanced techniques
of forgery and pas word-hacking. As an alternative, biometrics, such as palm-print, finger-print, face and ir s
being used for authentication and criminal identif cation [7]. The main advantage of biometrics is that these are not
prone to theft and los , and do not rely on the memory of
their users. Moreover, they do not change signif cantly over
time and it is dif icult for a person to alter own physiological
biometric or imitate that of another person's. Among
dif erent biometrics, in security ap lications with a scope of
col ecting digital identi y, the palm-prints are recently get ing
more at ention among researchers [3], [9]. Palm-print
recognit on is a complicated visual task even for humans. The primary dif iculty arises from the fact that dif erent
palm-print images of a particular person may vary largely, while those of dif erent persons may not neces arily vary
signif cantly. Moreover, some aspects of palm-prints, such
as variations in il umination, posit on, and scale, make the
recognit on task more complicated [6]. Palm-print
recognit on methods are based on extracting unique major
and minor line structures that remain stable throughout the
lifetime.
PROPOSED METHOD
For any type of biometric recognit on, the most important
task is to extract distinguishing features from
data, which directly dictates the recognit on ac uracy. In
comparison to person recognit on based on face or voice
biometrics, palm-print recognit on is very chal enging even
for a human being. For the case of palm-print recognit on, obtaining a signif cant feature space with respect to the
spatial variation in a palm-print image is very crucial. Moreover, a direct subjective cor espondence betwe n
palm-print features in the spatial domain and those in the
frequency domain is not very ap arent. In what fol ows, we
are going to demonstrate the proposed feature extraction
algorithm for palm-print recognit on, where spatial domain
local variation is extracted from frequency domain
transform.
5 CONCLUSIONS
In the proposed palm-print recognit on scheme, instead of
operating on the entire palm-print image at a time, dominant
DCT-based features are extracted separately from each of
the modules obtained by image-segmentation. It has be n
shown that because of modularization of the palm-print
image, the proposed dominant spectral features, that are
extracted from the sub-images, at ain bet er discriminating
capabil ties. The proposed feature extraction scheme is
shown to of er two-fold advantages. First, it can precisely
capture local variations that exist in the major and minor
lines of palm-print images, which plays an important role in
discriminating dif erent persons. Second, it util zes a very
low dimensional feature space for the recognit on task, which ensures lower computational burden. For the task of
clas if cation, an Euclidean distance based clas if er has
be n employed and it is found that, because of the quality
of the extracted features, such a simple clas if er can
provide a very satisfactory recognit on performance and
there is no ne d to employ any complicated clas if er. From
our extensive simulations on dif erent standard palm-print
databases, it has be n observed that he proposed method,
in comparison to some of the recent methods, provides
excel ent recognit on performance.