31-08-2012, 02:46 PM
FINGERPRINT VERIFICATION BASED ON IMAGE PROCESSING SEGMENTATION USING AN ONION ALGORITHM OF COMPUTATIONAL GEOMETRY
[attachment=33015]
INTRODUCTION
In this paper the problem of fingerprint verification via the Internet is
investigated. Specifically, the method that is used for the above purpose is
based on a traditional finger scanning technique, involving the analysis of
small unique marks of the finger image known as minutiae. Minutiae points
are the ridge endings or bifurcations branches of the finger image. The
relative position of these minutiae is used for comparison, and according
to empirical studies, two individuals will not have eight or more common
minutiae. [1,2]. A typical live-scan fingerprint will contain 30-40 minutiae.
Other systems analyze tiny sweat pores on the finger that, in the same way
as minutiae, are uniquely positioned. Furthermore, such methods may be
subject to attacks by hackers when biometric features are transferred via
Internet [3].
In our case we developed a method that addresses the problem of the
rotation and alignment of the finger position. The proposed method is
based on computational geometry algorithms (CGA). The advantages of
this method are based on a novel processing method using specific extracted
features, which may be characterized as unique to each person. These features
depend exclusively on the pixels brightness degree for the fingerprint
image, in contrast to traditional methods where features are extracted using
techniques such as edge and ridge - minutiae points detection. Specifically,
these feature express a specific geometric area (convex layer) in which the
dominant brightness value of the fingerprint ranges.
Pre-processing stage of CGA method
In this stage a fingerprint image, which is available from any of the known
image formats (tif, bmp, jpg, etc), is transformed into a matrix (a twodimensional
array) of pixels [5]. Consider, for example, the matrix of pixel
values of the aforementioned array. Then the brightness of each point is
proportional to the value of its pixel. This gives the synthesized image of
a bright square on a dark background. This value is often derived from
the output of an A/D converter. The matrix of pixels, i.e. the fingerprint
image, is usually square and an image will be described as N x N m-bit
pixels [6,7], where N is the number of points along the axes and m controls
the number of brightness values. Using m bits gives a range of 2 m values,
ranging from 0 to 2 m -1.
Verification stage based on Ratha’s algorithm
Fingerprint verification based on Ratha’s algorithm is a technique [11,12]
to assign a fingerprint into one of the several pre-specified types previously
described. Fingerprint verification can be viewed as a coarse level matching
of the fingerprints. An input fingerprint is first matched at a coarse level to
one of the pre-specified types and then, at a finer level, it is compared to the
subset of the database containing that type of fingerprint only. We have developed
an algorithm to classify fingerprints into five classes, namely, whorl,
right loop, left loop, arch, and tented arch. The algorithm separates the
number of ridges present in four directions (0 degree, 45 degree, 90 degree,
and 135 degree) by filtering the central part of a fingerprint with a bank
of Gabor filters. This information is quantized to generate a FingerCode,
which is used for classification [13,14].
RESULTS
In this experiment forty-eight (48) index-finger prints belonging to six (6)
individuals (6x8=48) and called A, B, C, D, E and F, were tested. More
specifically, each index-finger print of an individual was tested against the
other seven (7) in its group and the forty (40) prints of the other five
individuals. In total 2256 verification tests for each of the two methods
took place.
Statistical evaluation
As can be seen from the diagonal scores on in the above table (1) the correct
positive verification test score for the Ratha method, 156/168=0.93 or 93%
and for the CGA method is 165/168=0.98 or 98%. Furthermore, the correct
negative verification score for the Ratha algorithm is 933/940=0,99 or 99%
and the correct negative verification for the CGA method is 938/940 or
approximately 100%. In contrast, the false positive verification scores for
the Ratha method is 7% and for the CGA method 2%. At this point it is
to be noted that the false results of the Ratha method were yielded when
the tested image had variations for rotation reasons. On the other hand,
the CGA false results were yielded for those tested fingerprint specimens
that were not complete.
CONCLUSION
From the results of the experiment it is ascertained that the proposed
method, bearing in mind security considerations, can be used for accurate
and secure fingerprint verification purposes because the proposed feature
extraction is based on a specific area in which the dominant brightness value
of the fingerprint ranges. Moreover, the proposed method promisingly allows
very small false acceptance and false rejection rates, as it is based on
specific segmentation. It has to be noted that biometric applications will
gain universal acceptance in digital technologies only when the number of
false rejections / acceptances approach zero. The results of this comparison
showed that the proposed method yields correct positive and correct negative
verification scores greater than 99%.
[attachment=33015]
INTRODUCTION
In this paper the problem of fingerprint verification via the Internet is
investigated. Specifically, the method that is used for the above purpose is
based on a traditional finger scanning technique, involving the analysis of
small unique marks of the finger image known as minutiae. Minutiae points
are the ridge endings or bifurcations branches of the finger image. The
relative position of these minutiae is used for comparison, and according
to empirical studies, two individuals will not have eight or more common
minutiae. [1,2]. A typical live-scan fingerprint will contain 30-40 minutiae.
Other systems analyze tiny sweat pores on the finger that, in the same way
as minutiae, are uniquely positioned. Furthermore, such methods may be
subject to attacks by hackers when biometric features are transferred via
Internet [3].
In our case we developed a method that addresses the problem of the
rotation and alignment of the finger position. The proposed method is
based on computational geometry algorithms (CGA). The advantages of
this method are based on a novel processing method using specific extracted
features, which may be characterized as unique to each person. These features
depend exclusively on the pixels brightness degree for the fingerprint
image, in contrast to traditional methods where features are extracted using
techniques such as edge and ridge - minutiae points detection. Specifically,
these feature express a specific geometric area (convex layer) in which the
dominant brightness value of the fingerprint ranges.
Pre-processing stage of CGA method
In this stage a fingerprint image, which is available from any of the known
image formats (tif, bmp, jpg, etc), is transformed into a matrix (a twodimensional
array) of pixels [5]. Consider, for example, the matrix of pixel
values of the aforementioned array. Then the brightness of each point is
proportional to the value of its pixel. This gives the synthesized image of
a bright square on a dark background. This value is often derived from
the output of an A/D converter. The matrix of pixels, i.e. the fingerprint
image, is usually square and an image will be described as N x N m-bit
pixels [6,7], where N is the number of points along the axes and m controls
the number of brightness values. Using m bits gives a range of 2 m values,
ranging from 0 to 2 m -1.
Verification stage based on Ratha’s algorithm
Fingerprint verification based on Ratha’s algorithm is a technique [11,12]
to assign a fingerprint into one of the several pre-specified types previously
described. Fingerprint verification can be viewed as a coarse level matching
of the fingerprints. An input fingerprint is first matched at a coarse level to
one of the pre-specified types and then, at a finer level, it is compared to the
subset of the database containing that type of fingerprint only. We have developed
an algorithm to classify fingerprints into five classes, namely, whorl,
right loop, left loop, arch, and tented arch. The algorithm separates the
number of ridges present in four directions (0 degree, 45 degree, 90 degree,
and 135 degree) by filtering the central part of a fingerprint with a bank
of Gabor filters. This information is quantized to generate a FingerCode,
which is used for classification [13,14].
RESULTS
In this experiment forty-eight (48) index-finger prints belonging to six (6)
individuals (6x8=48) and called A, B, C, D, E and F, were tested. More
specifically, each index-finger print of an individual was tested against the
other seven (7) in its group and the forty (40) prints of the other five
individuals. In total 2256 verification tests for each of the two methods
took place.
Statistical evaluation
As can be seen from the diagonal scores on in the above table (1) the correct
positive verification test score for the Ratha method, 156/168=0.93 or 93%
and for the CGA method is 165/168=0.98 or 98%. Furthermore, the correct
negative verification score for the Ratha algorithm is 933/940=0,99 or 99%
and the correct negative verification for the CGA method is 938/940 or
approximately 100%. In contrast, the false positive verification scores for
the Ratha method is 7% and for the CGA method 2%. At this point it is
to be noted that the false results of the Ratha method were yielded when
the tested image had variations for rotation reasons. On the other hand,
the CGA false results were yielded for those tested fingerprint specimens
that were not complete.
CONCLUSION
From the results of the experiment it is ascertained that the proposed
method, bearing in mind security considerations, can be used for accurate
and secure fingerprint verification purposes because the proposed feature
extraction is based on a specific area in which the dominant brightness value
of the fingerprint ranges. Moreover, the proposed method promisingly allows
very small false acceptance and false rejection rates, as it is based on
specific segmentation. It has to be noted that biometric applications will
gain universal acceptance in digital technologies only when the number of
false rejections / acceptances approach zero. The results of this comparison
showed that the proposed method yields correct positive and correct negative
verification scores greater than 99%.