A NEW CODING METHOD IN PATTERN RECOGNITION FINGERPRINT IMAGE USING VECTOR QUANTIZATION
1 like317 views
The document presents a new coding method for fingerprint image recognition using vector quantization, aimed at addressing challenges in matching poor quality images and large fingerprint databases. It discusses the limitations of current fingerprint recognition systems, highlights the importance of a unique vector representation for effective matching, and outlines the proposed algorithm's mechanism involving image compression. The approach is expected to significantly enhance identification speed and reduce memory requirements compared to traditional methods.
![International Journal of Technical Research and Applications e-ISSN: 2320-8163,
www.ijtra.com Volume 2, Issue 3 (May-June 2014), PP. 32-34
32 | P a g e
A NEW CODING METHOD IN PATTERN
RECOGNITION FINGERPRINT IMAGE USING
VECTOR QUANTIZATION
NAGHAM J. SHUKUR
M.Sc in Space technologies and Communicatinon
Unit of Applied Sciences Researchs, School of Applied Sciences, University of Technology,
Baghdad, Iraq
Abstract— Fingerprints are imprints formed by friction
ridges of the skin and thumbs. They have long been used for
identification because of their immutability and individuality.
Immutability refers to the permanent and unchanging character
of the pattern on each finger. Individuality refers to the
uniqueness of ridge details across individuals; the probability
that two fingerprints are alike is about 1 in 1.9x1015. In despite of
this improvement which is adopted by the Federal Bureau of
Investigation (FBI), the fact still is “The larger the fingerprint
files became, the harder it was to identify somebody from their
fingerprints alone. Moreover, the fingerprint requires one of the
largest data templates in the biometric field”. The finger data
template can range anywhere from several hundred bytes to over
1,000 bytes depending upon the level of security that is required
and the method that is used to scan one's fingerprint. For these
reasons this work is motivated to present another way to tackle
the problem that is relies on the properties of Vector
Quantization coding algorithm.
Key words— Vector Quantization, Fingerprint, Vector
Quantization coding algorithm
I. INTRODUCTION
Nowadays, fingerprint recognition is one of the most
important biometric technologies based on fingerprint
distinctiveness, persistence and ease of acquisition. Although
there are many real applications using this technology, its
problems are still not fully solved, especially in poor quality
fingerprint images and when low-cost acquisition devices with
a small area are adopted.
Fig.1 Fingerprint Recognition process
In fingerprint recognition process, the important step which
affects on system accuracy is matching between template and
query fingerprint. Many solutions are
designed to increase this step’s accuracy ([1], [2], [3], [4],
[5], [6]). Fingerprint imaging technology looks to capture or
read the unique pattern of lines on the tip of one's finger. These
unique patterns of lines can either be in a loop, whorl, or arch
pattern [7]. There are several methods in accomplishing the
process of identifying one's fingerprint [8]. The two major
applications of fingerprint recognition are
fingerprint verification and fingerprint identification.
Verification is known as one-to-one fingerprint matching,
whereas, fingerprint identification is known as one-to-many
matching. Verification is used for access applications and
identification is used for investigation purposes. Many
successful approaches have been presented for verification
applications, where the identification field is still open
challenging area because of many problems. The main
common problem that the verification faces is the huge
database. This problem can be classified into two categories:
first is
establishing identify based on single print i.e. comparing a
latent print from the scene of a crime with prints from a file.
Second is establishing identify based on a set of ten fingerprint
records i.e. record [9]. After arrest or employment are
requirements. For reasons discussed in [10], the problem of the
first category is extremely difficult, whereas the second one has
greater quantity of information available from ten-print file
searching. The traditional systems have sequential structure
which suffers from a problem that the error propagates and
information is blocked which means that the second stage can
not use the information available at the first stage but it may
suffer from the errors caused by the first stage. Moreover,
conventionally, the matching process may use the following
features: ridges or valleys (or both), minutia i.e. its type,
location and direction, and global matching. Common Features
are shown in Table 1 and Figure 1 respectively.
Table 1. Common Features for Matching Process
Ridges Various patterns across
fingerprint
Valleys Spaces between ridges
Minutia Type (ending,
bifurcation)Location
(x,y)
Direction
Global
Patterns
Arch LoopWhorl](https://image.slidesharecdn.com/ijtra140319-151003062358-lva1-app6891/85/A-NEW-CODING-METHOD-IN-PATTERN-RECOGNITION-FINGERPRINT-IMAGE-USING-VECTOR-QUANTIZATION-1-320.jpg)
![International Journal of Technical Research and Applications e-ISSN: 2320-8163,
www.ijtra.com Volume 2, Issue 3 (May-June 2014), PP. 32-34
33 | P a g e
Figure 2. Common global Patterns (above), common
Minutia
This work proposes new feature that can be used for
matching. The proposed feature is a vector generated after the
fingerprint image is compressed by Vector Quantization
Coding approach. This vector is uniquely representing the
entire image. Thus, it can
be effectively used for matching purposes. The proposed
algorithm is illustrated in section 3 and tested in section 4.
Section 5 concludes this work and gives some
recommendations. Before all, brief introduction to Vector
Quantization coding is given in next section.
2. Vector Quantization Coding Algorithm:
It uses a codebook that contains the pixel patterns with
the corresponding index on each of them. The main idea of VQ
in image coding is then to represent arrays of pixels by an
index in the codebook. In this way, compression is achieved
since the size of the index is usually a small fraction of that of
the block of pixels [11].
3. PROPOSED APPROCH:
A. Mechanism Of The proposed system :
In Figure 3. depicts the general mechanism of the
proposed system. After the fingerprint image is fed to the
system, it will be converted to binary image as a preprocessing
stage. Then the binary image is compressed using Vector
Quantization coding approach. The first stage in image vector
quantization is the image formation. The image data are first
partitioned into a set of vectors. A large number of vectors
from various images are then used to form a training set. The
training set is used to generate a codebook, normally using an
iterative clustering algorithm. The quantization or coding step
involves searching each input vector for the closest codeword
in the codebook. Then the corresponding index of the selected
codeword is coded and transmitted to the decoder. At the
decoder, the index is decoded and converted to the
corresponding vector with the same codebook as at the encoder
by look-up table. Thus, the design decisions in implementing
image vector quantization include [12]:
1. vector formation;
2. training set generation;
3. code book generation; and
4. quantization.
Figure 3. Flowchart for the proposed algorithm
the distance between the entire vector and the
corresponding vector in codebook is equal to zero, but because
of the entire image has not always the exact appearance to the
image that captured during the stage of building the codebook,
the distance may be greater. However, the shortest distance can
be used to point to the corresponding vector and then
fingerprint is identified.
Procedure Of Applying Proposed System:
The two important components for the system shown above
are the Vector Quantization coding compression stage and the
codebook. At compression stage the entire image which is
binary matrix is compressed using Vector Quantization coding
algorithm to produce a vector of almost 1/16 image size. For
instance, for the fingerprint image of size 256x256 or 65536
which uses 65536 bytes, the generated vector length is 4096
that is using 8192 bytes. Therefore, the memory size is
decreased by 1/8, from 65536bytes for original image to
8192bytes for corresponding vector. To build the codebook,
same procedure can be followed, by converting the database
(fingerprint images) to code vectors and assigning each one to
its corresponding owner. Thus the codebook contains vectors
rather than images. For instance, if we have 1 million finger
images for 1 million persons, this database is traditionally using
256x256x1000000 = 655360000000 bytes ~ = 154
Gigabytes, their corresponding vectors use only
20720x1000000 bytes ~= 19 Gigabytes. These simple
calculations lead to conclude that the identification rate has
gained many benefits such as high speed as well as low False
Acceptance Rate (FAR) / False Rejection Rate (FRR).
REFERENCES
[1] A. Jain, L. Hong, and R. Bolle, “On-line fingerprint
verification,” IEEE Trans. Pattern Anal. Machine Intell, vol.
19, pp 302–314, 1997. X. Jiang, W. Y. Yau, "Fingerprint
minutiae matching based on the local and global structures",
ICPR2000, vol. 2, 2000, pp 1042-1045.](https://image.slidesharecdn.com/ijtra140319-151003062358-lva1-app6891/85/A-NEW-CODING-METHOD-IN-PATTERN-RECOGNITION-FINGERPRINT-IMAGE-USING-VECTOR-QUANTIZATION-2-320.jpg)
![International Journal of Technical Research and Applications e-ISSN: 2320-8163,
www.ijtra.com Volume 2, Issue 3 (May-June 2014), PP. 32-34
34 | P a g e
[2] A.K. Jain, S. Prabhakar, L. Hong, S. Pankanti, "Filterbank-
Based Fingerprint Matching", Image Processing, vol. 9, 2000, pp
846-859.
[3] K. Ito, A. Morita, T. Aoki, H. Nakajima, K. Kobayashi, T.
Higuch, "A Fingerprint Recognition Algorithm Combining
Phase-Based Image Matching and Feature-Based Matching",
International Conference on Biometrics, vol. 3832, 2006, pp
316–325.
[4] 4. L. Sha, F. Zhao, X. Tang, "Minutiae-based Fingerprint
Matching Using Subset Combination", The 18th
International
Conference on Pattern Recognition -ICPR'06, vol. 4, 2006, pp
566-569.
[5] X. Jiang, W. Y. Yau, "Fingerprint minutiae matching based on
the local and global structures", ICPR2000, vol. 2, 2000, pp.
1042-1045
[6] Xuejun Tan, Bir Bhanu, "Fingerprint matching by genetic
algorithms", Pattern Recognition, vol. 39, 2006, pp. 465 –477.
[7] Randell, N., Biometric basics, PC Magazine, April 1999, pp.
193-202.
[8] The Center for Unified Biometrics and Sensors
(CUBS), Fingerprint recognition, retrieved August 13, 2007:
http://www.cubs.buffalo.edu/fingerrecog.sht ml. Accessed
January 2008
[9] Abdurazzag Ali Aburas and Salem Ali Reheil, “New Promising
Tool for Off Line Arabic Handwritten Character Recognition
Based On JPEG2000 Image Compression”, IEEE-3rd
International Conference on Information & Communication
Technologies: From theory to Applications (ICTTA’08),
Damascus-Syria, April 47-11 2008
[10] McMahon, D.H.; Johnson, G.L.; Teeter, S.L.; Whitney, C.G.
(April 1975) A Hybrid Optical Computer Processing Technique
for Fingerprint Identification, IEEE Transactions on Computers
Volume C-24, Issue 4, Page(s):358 – 369
[11] Loben, C.: Introduction to Image Compression,
http://www/image.cityu.edu.hkl~/thesis/node9.html; 1998.
[12] Shi, Y.Q. and Sun, H.: Image and Video Compression for
Multimedia Engineering: Fundamentals, Algorithms and
Standards (1st ed.). CRC Press LLC, New York; 2000](https://image.slidesharecdn.com/ijtra140319-151003062358-lva1-app6891/85/A-NEW-CODING-METHOD-IN-PATTERN-RECOGNITION-FINGERPRINT-IMAGE-USING-VECTOR-QUANTIZATION-3-320.jpg)
A NEW CODING METHOD IN PATTERN RECOGNITION FINGERPRINT IMAGE USING VECTOR QUANTIZATION
- 1. International Journal of Technical Research and Applications e-ISSN: 2320-8163, www.ijtra.com Volume 2, Issue 3 (May-June 2014), PP. 32-34 32 | P a g e A NEW CODING METHOD IN PATTERN RECOGNITION FINGERPRINT IMAGE USING VECTOR QUANTIZATION NAGHAM J. SHUKUR M.Sc in Space technologies and Communicatinon Unit of Applied Sciences Researchs, School of Applied Sciences, University of Technology, Baghdad, Iraq Abstract— Fingerprints are imprints formed by friction ridges of the skin and thumbs. They have long been used for identification because of their immutability and individuality. Immutability refers to the permanent and unchanging character of the pattern on each finger. Individuality refers to the uniqueness of ridge details across individuals; the probability that two fingerprints are alike is about 1 in 1.9x1015. In despite of this improvement which is adopted by the Federal Bureau of Investigation (FBI), the fact still is “The larger the fingerprint files became, the harder it was to identify somebody from their fingerprints alone. Moreover, the fingerprint requires one of the largest data templates in the biometric field”. The finger data template can range anywhere from several hundred bytes to over 1,000 bytes depending upon the level of security that is required and the method that is used to scan one's fingerprint. For these reasons this work is motivated to present another way to tackle the problem that is relies on the properties of Vector Quantization coding algorithm. Key words— Vector Quantization, Fingerprint, Vector Quantization coding algorithm I. INTRODUCTION Nowadays, fingerprint recognition is one of the most important biometric technologies based on fingerprint distinctiveness, persistence and ease of acquisition. Although there are many real applications using this technology, its problems are still not fully solved, especially in poor quality fingerprint images and when low-cost acquisition devices with a small area are adopted. Fig.1 Fingerprint Recognition process In fingerprint recognition process, the important step which affects on system accuracy is matching between template and query fingerprint. Many solutions are designed to increase this step’s accuracy ([1], [2], [3], [4], [5], [6]). Fingerprint imaging technology looks to capture or read the unique pattern of lines on the tip of one's finger. These unique patterns of lines can either be in a loop, whorl, or arch pattern [7]. There are several methods in accomplishing the process of identifying one's fingerprint [8]. The two major applications of fingerprint recognition are fingerprint verification and fingerprint identification. Verification is known as one-to-one fingerprint matching, whereas, fingerprint identification is known as one-to-many matching. Verification is used for access applications and identification is used for investigation purposes. Many successful approaches have been presented for verification applications, where the identification field is still open challenging area because of many problems. The main common problem that the verification faces is the huge database. This problem can be classified into two categories: first is establishing identify based on single print i.e. comparing a latent print from the scene of a crime with prints from a file. Second is establishing identify based on a set of ten fingerprint records i.e. record [9]. After arrest or employment are requirements. For reasons discussed in [10], the problem of the first category is extremely difficult, whereas the second one has greater quantity of information available from ten-print file searching. The traditional systems have sequential structure which suffers from a problem that the error propagates and information is blocked which means that the second stage can not use the information available at the first stage but it may suffer from the errors caused by the first stage. Moreover, conventionally, the matching process may use the following features: ridges or valleys (or both), minutia i.e. its type, location and direction, and global matching. Common Features are shown in Table 1 and Figure 1 respectively. Table 1. Common Features for Matching Process Ridges Various patterns across fingerprint Valleys Spaces between ridges Minutia Type (ending, bifurcation)Location (x,y) Direction Global Patterns Arch LoopWhorl
- 2. International Journal of Technical Research and Applications e-ISSN: 2320-8163, www.ijtra.com Volume 2, Issue 3 (May-June 2014), PP. 32-34 33 | P a g e Figure 2. Common global Patterns (above), common Minutia This work proposes new feature that can be used for matching. The proposed feature is a vector generated after the fingerprint image is compressed by Vector Quantization Coding approach. This vector is uniquely representing the entire image. Thus, it can be effectively used for matching purposes. The proposed algorithm is illustrated in section 3 and tested in section 4. Section 5 concludes this work and gives some recommendations. Before all, brief introduction to Vector Quantization coding is given in next section. 2. Vector Quantization Coding Algorithm: It uses a codebook that contains the pixel patterns with the corresponding index on each of them. The main idea of VQ in image coding is then to represent arrays of pixels by an index in the codebook. In this way, compression is achieved since the size of the index is usually a small fraction of that of the block of pixels [11]. 3. PROPOSED APPROCH: A. Mechanism Of The proposed system : In Figure 3. depicts the general mechanism of the proposed system. After the fingerprint image is fed to the system, it will be converted to binary image as a preprocessing stage. Then the binary image is compressed using Vector Quantization coding approach. The first stage in image vector quantization is the image formation. The image data are first partitioned into a set of vectors. A large number of vectors from various images are then used to form a training set. The training set is used to generate a codebook, normally using an iterative clustering algorithm. The quantization or coding step involves searching each input vector for the closest codeword in the codebook. Then the corresponding index of the selected codeword is coded and transmitted to the decoder. At the decoder, the index is decoded and converted to the corresponding vector with the same codebook as at the encoder by look-up table. Thus, the design decisions in implementing image vector quantization include [12]: 1. vector formation; 2. training set generation; 3. code book generation; and 4. quantization. Figure 3. Flowchart for the proposed algorithm the distance between the entire vector and the corresponding vector in codebook is equal to zero, but because of the entire image has not always the exact appearance to the image that captured during the stage of building the codebook, the distance may be greater. However, the shortest distance can be used to point to the corresponding vector and then fingerprint is identified. Procedure Of Applying Proposed System: The two important components for the system shown above are the Vector Quantization coding compression stage and the codebook. At compression stage the entire image which is binary matrix is compressed using Vector Quantization coding algorithm to produce a vector of almost 1/16 image size. For instance, for the fingerprint image of size 256x256 or 65536 which uses 65536 bytes, the generated vector length is 4096 that is using 8192 bytes. Therefore, the memory size is decreased by 1/8, from 65536bytes for original image to 8192bytes for corresponding vector. To build the codebook, same procedure can be followed, by converting the database (fingerprint images) to code vectors and assigning each one to its corresponding owner. Thus the codebook contains vectors rather than images. For instance, if we have 1 million finger images for 1 million persons, this database is traditionally using 256x256x1000000 = 655360000000 bytes ~ = 154 Gigabytes, their corresponding vectors use only 20720x1000000 bytes ~= 19 Gigabytes. These simple calculations lead to conclude that the identification rate has gained many benefits such as high speed as well as low False Acceptance Rate (FAR) / False Rejection Rate (FRR). REFERENCES [1] A. Jain, L. Hong, and R. Bolle, “On-line fingerprint verification,” IEEE Trans. Pattern Anal. Machine Intell, vol. 19, pp 302–314, 1997. X. Jiang, W. Y. Yau, "Fingerprint minutiae matching based on the local and global structures", ICPR2000, vol. 2, 2000, pp 1042-1045.
- 3. International Journal of Technical Research and Applications e-ISSN: 2320-8163, www.ijtra.com Volume 2, Issue 3 (May-June 2014), PP. 32-34 34 | P a g e [2] A.K. Jain, S. Prabhakar, L. Hong, S. Pankanti, "Filterbank- Based Fingerprint Matching", Image Processing, vol. 9, 2000, pp 846-859. [3] K. Ito, A. Morita, T. Aoki, H. Nakajima, K. Kobayashi, T. Higuch, "A Fingerprint Recognition Algorithm Combining Phase-Based Image Matching and Feature-Based Matching", International Conference on Biometrics, vol. 3832, 2006, pp 316–325. [4] 4. L. Sha, F. Zhao, X. Tang, "Minutiae-based Fingerprint Matching Using Subset Combination", The 18th International Conference on Pattern Recognition -ICPR'06, vol. 4, 2006, pp 566-569. [5] X. Jiang, W. Y. Yau, "Fingerprint minutiae matching based on the local and global structures", ICPR2000, vol. 2, 2000, pp. 1042-1045 [6] Xuejun Tan, Bir Bhanu, "Fingerprint matching by genetic algorithms", Pattern Recognition, vol. 39, 2006, pp. 465 –477. [7] Randell, N., Biometric basics, PC Magazine, April 1999, pp. 193-202. [8] The Center for Unified Biometrics and Sensors (CUBS), Fingerprint recognition, retrieved August 13, 2007: http://www.cubs.buffalo.edu/fingerrecog.sht ml. Accessed January 2008 [9] Abdurazzag Ali Aburas and Salem Ali Reheil, “New Promising Tool for Off Line Arabic Handwritten Character Recognition Based On JPEG2000 Image Compression”, IEEE-3rd International Conference on Information & Communication Technologies: From theory to Applications (ICTTA’08), Damascus-Syria, April 47-11 2008 [10] McMahon, D.H.; Johnson, G.L.; Teeter, S.L.; Whitney, C.G. (April 1975) A Hybrid Optical Computer Processing Technique for Fingerprint Identification, IEEE Transactions on Computers Volume C-24, Issue 4, Page(s):358 – 369 [11] Loben, C.: Introduction to Image Compression, http://www/image.cityu.edu.hkl~/thesis/node9.html; 1998. [12] Shi, Y.Q. and Sun, H.: Image and Video Compression for Multimedia Engineering: Fundamentals, Algorithms and Standards (1st ed.). CRC Press LLC, New York; 2000