International Journal of Scientific & Engineering Research, Volume 6, Issue 5, May-2015 129

ISSN 2229-5518

Proposal Fingerprint Recognition Regimes Development Based on Minutiae Matching Hany Hashem Ahmed, Hamdy M. Kelash, Maha S. Tolba, Mohammed Badawy.

Computer Science and Engineering, Faculty of Electronic Engineering, Menoufia University, Egypt.

Abstract— Fingerprint recognition is one of the oldest and most popular biometric technology and it is used in criminal investigations, civilian, commercial applications, and so on. Fingerprint matching is the process used to determine whether the two sets of fingerprints details come from the same finger or not. This work focuses on feature extraction and minutiae matching stage. There are many matching techniques used for fingerprint recognition systems such as minutiae based matching, pattern based matching, Correlation based matching, and image based matching. Two fingerprint recognition regimes have been developed based on minutiae matching, the first one is: Artificial Neural Network based on Minutiae Distance Vector (ANN-MDV), while the other one is: Artificial Neural Network based on Principle Component Analysis (ANN-PCA). It is observed that the recognition rate is increased and return better result. A comparative study between the 2-developed system is done based on average recognition time (ART), and the accuracy of the recognition system. The experimental results are done on FVC2002 database. These results show that the accuracy of ANN-MDV system is approximately equal to

91%, and the accuracy of ANN-PCA system is approximately equal to 98%. Therefore ANN-PCA is the best recognition system accuracy. Also the experimental results show that ART for ANN-MDV (equal to 0.251) is slightly better than ANN-PCA (equal to 0.275).

Index Terms— Fingerprint Recognition, image enhancement, FDCT, Minutiae Distance Vector, ANN, BPN, PCA, ART.

—————————— ——————————

1 INTRODUCTION

A fingerprint consists of a pattern of ridges (lines across fingerprints) and valleys (spaces between ridges) in a finger. The pattern of the ridges and valleys is unique, permanent for each individual, and remains unchanged over a lifetime. Mi- nutiae (fingerprint features) are formed from the local discon- tinuities in ridge flow pattern. These minutiae have the re- quired features that are used in fingerprint recognition system. There are many types of minutiae like Bifurcation, Termina- tion, Lake, Spur, Crossover, dot, bridge, trifurcation, island, and singular points (core & delta). The considered types of extracted features used in this paper for fingerprint recogni- tion are ridge bifurcation point, ridge termination point, core point, and delta point as seen in Fig. 1 [1, 2].

(a) Bifurcation (b) Termination

(c) Core & Delta points

Fig. 1 Types of minutiae

The technique used here for fingerprint recognition is
based on minutiae matching. The fingerprint recognition sys-
tem is a comparison between the input fingerprint image and the template fingerprint image stored previously in a data-
base. The main purpose of this work is to develop a new tech- nique for fingerprint recognition system that return an excel- lent results to query the input fingerprint image from the da-
tabase in an acceptable response time. There are a large num- ber of techniques that are being used for fingerprint recogni- tion systems; one of them is artificial neural network (ANN). ANN is an efficient method for prediction and recognition. There are many types of network such as Perceptron , feed forward back propagation network, radial basis network, Hopfield recurrent network, pattern recognition network, etc.., in this paper feed forward back propagation network has been used for the developed system. In this paper for the 2- proposed systems, back propagation network is the best net- work in training and return relevant results. Two fingerprint recognition systems have been proposed and developed based on ANN, the first one system is ANN based on minutiae dis- tance vector (ANN-MDV), and the second is ANN based on Principle component analysis (ANN-PCA). The rest of this paper is organized as follow: Section 2 discusses the principle of ANN, advantages of ANN, MDV description and explains the work of PCA. Section 3 shows the block diagrams of the 2- developed systems and discusses the main stages of each one. Section 4 shows the experimental results and examines the recognition systems. Section 5 presents a comparative analysis between the 2-developed recognition regimes, and introduces a comparison tables. The last section gives a brief summary, conclusion, and represents short notes for future work.

2 RELATED WORK

This section presents a brief description about neural net- work, principle component analysis, and minutiae distance vector.

2.1 Minutiae Distance Vector (MDV)

Minutiae Distances (MDs) are the distances between the refer- ence point (core point) and all minutiae points (bifurcation, termination, and delta points). Minutiae Distance Vector (MDV) can be calculated by sorting these estimated distances (MDs) in ascending form and put it in one vector.

IJSER © 2015 http://www.ijser.org

International Journal of Scientific & Engineering Research, Volume 6, Issue 5, May-2015 130

ISSN 2229-5518

2.2 Principle Component Analysis (PCA)

Due to great difficulties in determining similarities and differ- ences between data arising from large patterns of data, there- fore we use PCA to solve this problem. PCA is an efficient and a powerful tool for analyzing data patterns. Another im- portant feature of PCA is the ability to compress data by re- ducing the number of dimensions without losing much infor- mation. Finally PCA could be defined as a statistical proce- dure (variance, covariance, mean, eigenvector….etc.) used to convert patterns of data with related variables into a set of values of non-related variables called principal components (PC), these PCs are always less than or equal to the original related variables [3, 4].

2.3 Artificial Neural Network (ANN)

Definition: ANN is an information processing system that has certain performance characteristics similar to biological neural network. Description: neural network consists of large number of simple processing units called neurons or nodes. Each node is connected to the other nodes by direct communication links. Each link has an associated weight. The weight contains in- formation used by the network to solve the problem [5, 6]. ANN can be used to store and query data or pattern, classify pattern and find solution to constrained optimization prob- lems. Fig. 2 shows a simple neuron [7].

Fig. 2 Simple neuron or node.

The function of simple neuron can be described by (1).

(a) Single layer

(b)Multilayer

(c) Recurrent

𝑜𝑢𝑡𝑝𝑢𝑡 = �

1 𝑖𝑓 ∑𝑛

0 𝑖𝑓 ∑𝑛

𝑤𝑖 . 𝑥𝑖

𝑤𝑖 . 𝑥𝑖

≥ 𝑇

< 𝑇

(1)

Fig. 3 Neural network architectures

2- Training or learning types: are methods to estimate the
Where T is threshold, Xi is the input to neuron, Wi is the
weight, and n is the number of inputs. The neural network is characterized by:
1- Architecture: is a pattern of connections among the neurons
(arrangement of neurons into layer) [7]. As shown in Fig. 3 there are three architecture types as follow:
a. Single layer feed forward network: has one layer of con- nection weights
b. Multilayer feed forward network: is a network with one
weights on the connections. Learning types:
a. Supervised learning: each input vector of the network
has an associated target output vector. Every learning cy-
cle the error (difference between the actual and desired output) is used to adjust the weights.
b. Unsupervised learning: the input vectors are provided to the network, but with no associated target vectors. The weights are adjusted so that the similar input vectors are
assigned to the same output.
or more layer (called hidden layer) between the input 3- Activation functions: determine how the output of the neu-
units and the output units.
c. Recurrent network: there are closed loop signal paths
from a unit back to itself.
ron will be calculated [7]. Some of the activation functions are seen in Fig. 4.
a. Binary step function (with threshold t). b. Bipolar binary function.

c. Sigmoid function.

𝑓(𝑥) = �1 𝑖𝑓 𝑥 ≥ 𝑡

0 𝑖𝑓 𝑥 < 𝑡

IJSER © 2015 http://www.ijser.org

(a) Binary step function with threshold t.

(2)

International Journal of Scientific & Engineering Research, Volume 6, Issue 5, May-2015 131

ISSN 2229-5518

𝑓(𝑖𝑛) = � 1 𝑖𝑓 𝑖𝑛 ≥ 0

−1 𝑖𝑓 𝑖𝑛 < 0

(b) Bipolar binary function.

(3)

1

𝑓 (𝑖𝑛) ≜

1 + exp(−𝑖𝑛)

(c) Sigmoid function.

(4)

Fig. 4 Activation functions.

3 PROPOSED SYSTEM

Tow systems are developed as follow:

First system: Fingerprint recognition system using ANN based

on MDV.

Second system: Fingerprint recognition system using ANN

based on PCA.

3.1 Fingerprint recognition system using ANN based on MDV

The design of the first developed system is shown in Fig. 5.

Fig. 5 Fingerprint recognition ANN-MDV

The stages of the algorithm are fingerprint image acquisition,

preprocessing, features extraction, minutiae distance vector, train-

ing the network, testing the network, matching the input finger- print with stored fingerprint images in data base, finally esti- mate the result.

1) Preprocessing stage
The algorithm of preprocessing stage is shown in Fig. 6.

Input fingerprint image

Fast curvelet denoising Segmentation Normalization Orientation Estimation

Frequency Estimation

Gabor filtering

Thinning

Enhanced fingerprint image to feature extraction step

Fig. 6 Preprocessing stage

Practically the input fingerprint image may be noisy and cor- rupted due to environmental factors or body condition of the
user. So that it is very important to do some preprocessing steps on the input fingerprint image in order to improve the clarity of ridge structure and increase the performance of mi- nutiae extraction algorithm. Therefore the main purpose of preprocessing stage is to enhance and preparing the input fin- gerprint image for next stage. The steps of preprocessing stage are:
Step 1. Fast curvelet denoising: In this step fast curvelet is
used to eliminate different kinds of noise such as random
noise, salt noise, speckle noise and Gaussian noise form
fingerprint images. The algorithm of fingerprint image denoising done by:
1- Compute all thresholds of curvelet which will be ap- plied to image curvelet coefficient.
2- Normalize the curvelet coefficient.
3- Perform warping Fast Discrete Curvelet Transform (FDCT) to the noisy image and transfer it from spatial domain to curvelet domain.
4- Apply the computed threshold in step1 to the curvelet coefficients.
5- Apply inverse fast discrete curvelet transform to the re-
sult in order to transfer the image from curvelet domain to spatial domain (original state) [8, 9].
Step 2. Segmentation: Is the process of separating the fore- ground regions (contain fingerprint information which is called the Region of Interest ROI) from the background
regions (noisy area) in the fingerprint image [10].

IJSER © 2015 http://www.ijser.org

International Journal of Scientific & Engineering Research, Volume 6, Issue 5, May-2015 132

ISSN 2229-5518

Step 3. Normalization: The main purpose of normalization is to reduce the differences in the grey level values and en- hance the contrast of the fingerprint image, so that the
central pixel (P) equal to 1 and has exactly 3 one-value neigh- bors, then the central pixel is a ridge bifurcation point. The CN value can be estimated by (5).
ridges and valleys of the normalized image can be easily

CN = 0.5 ∑8

|Pi − Pi+1 | , P9 = P1

distinguished [1].
Step 4. Orientation estimation: In this step the dominant di-
rection of the ridges and valleys in the fingerprint image
is estimated [11].
Step 5. Frequency estimation: The local ridge frequency is defined as the frequency of the ridge and valley structures in a local neighborhood along a direction orthogonal to the local ridge orientation. The ridge frequency is also varying slowly and hence it is computed only once for each non-overlapping block of the image [12].
Step 6. Gabor filter: The purpose of this step is to remove noise and preserve the ridge and valley structures [13].
Step 7. Thinning (skeletonization): After the thinning process is applied on the fingerprint image, the ridge width be- comes one pixel size (skeleton fingerprint image). The
ridge thinning process make the features extraction and marking minutiae points are very easy [14].
2) Feature extraction stage
Fingerprint image contains a lot of minutiae such as ridge
termination, ridge bifurcation, short ridge, island, crossing

point, delta point, and core point. But the interested and most important features considered in this paper are ridge termina- tion, ridge bifurcation, and singular points as shown in Fig. 7. Core Point is the topmost point on the innermost upwardly curving ridgeline (approximately center of the fingerprint). Core point is considered as reference point for reading and classifying the fingerprint image. Delta point is defined as the point on the first bifurcation, meeting of two ridges, fragmen- tary ridge, abrupt ending ridge dot, or any point on a ridge at or in front of and nearest to the center of the divergence of the type lines. Poincare index algorithm is used to extract the sin- gular points core and delta.

Fig. 7 A core and a delta singularity in a right loop fingerprint [1].

The algorithm used for extracting features from fingerprint image is Crossing Number (CN) which consider (3x3) pixel window as shown in Fig. 8.

Fig. 8 A 3 x3 window is placed on a binary image, pixel P with its 8 neigh- boring points (P1, P2……P8).

If the central pixel (P) is 1 and has only 1 one-value neighbour,
then the central pixel is a ridge termination point. Also if the

i=1 (5)

Where Pi is the pixel value in the neighborhood of P. The pix-
els can be classified according to the value of CN as shown in table 1 [10, 15].

TABLE1 MINUTIAE CLASSIFICATION

Remove false minutiae: removing false minutiae is very im- portant step for the accuracy of fingerprint recognition system. The algorithm of removing false minutiae as follows:

1. First we calculate the average distance "D" between 2- parallel neighboring ridges and suppose D as threshold for false minutiae.
2. If the distance between ending point and bifurcation point
is less than D and the two minutiae are in the same ridge,
then remove both of them.
3. If the distance between two bifurcations is less than D and they are in the same ridge, remove the two bifurcations.
4. If the distance between 2-terminations is less than D, re- move the two terminations.
5. If 2-termination points are within a distance D and their directions are coincident with a small angle variation,
then the 2-termination points are considered as false mi- nutia and are removed.

Direction and angle of correct minutiae: As discussed before the important minutiae points are ridge ending "CN = 1" and ridge bifurcation "CN = 3", therefore the direction and angle of

these minutiae are very important. The 8 directions (N, S, W, E, NE, NW, SE, and SW) can be determined by the following pseudo code:
% for ridge ending point
If CN = 1 then
If P1 = 1 then direction = W If P3 = 1 then direction = S If P7 = 1 then direction = N If P5 = 1 then direction = E
If P4 = 1 then direction = SE If P2 = 1 then direction = SW
If P6 = 1 then direction = NE If P8 = 1 then direction = NW End if
% for ridge bifurcation point
If CN = 3 then
If P1 and P3 and P7 = 1 then direction = W If P1 and P3 and P5 = 1 then direction = S If P1 and P7 and P5 = 1 then direction = N If P3 and P5 and P7 = 1 then direction = E
If P4 and P3 and P5 = 1 then direction = SE
If P3 and P2 and P1 = 1 then direction = SW If P3 and P5 and P6 = 1 then direction = NE If P4 and P8 and P5 = 1 then direction = NW

IJSER © 2015 http://www.ijser.org

International Journal of Scientific & Engineering Research, Volume 6, Issue 5, May-2015 133

ISSN 2229-5518

End if [16].
3) Minutiae Distance Vector (MDV) stage
MDV is the input vector to the neural network. This vector can
be obtained and formed by calculating the distance between each minutiae coordinates & the reference point (core point), and then these distances are sorted in ascending form. After MDVs are formed for all fingerprints images, these vectors will be saved and stored in a database and will be input to the neural network.
4) ANN Training stage
Practically, the best artificial neural network type for finger-
print recognition system is feed forward back propagation
network. Structure of feed forward back propagation network
is shown in Fig. 9.
responding to class P101, the second person is corresponding to class P201, and so on. Then these data will be stored in data base for matching process. Now the network is ready to iden- tify the fingerprint image. When the input fingerprint image enters to the system, minutiae are transferred to a vector, and then the network simulates this vector and gives the result.

3.2 Fingerprint recognition system using ANN based on PCA


The algorithm of this system is shown in Fig. 10. There is only one big difference between the 2-developed systems in which PCA is used instead of MDV in stage 3.

Fig. 9 Structure of back propagation neural network [7]

Once the network structure has been created, the training phase is ready to begin. The input data (MDVs) are divided into training set 70%, testing set 15%, and validation set 15%. The supervised training technique has been chosen to train feed forward back propagation network. Neural network units (neurons) are trained with Scaled conjugate gradient back propagation algorithm called (trainscg). Performance function of the developed network is Mean squared error with regular- ization performance function (msereg). The activation transfer function used in the network is hyperbolic tangent sigmoid transfer function (tansig). The training algorithm as follow [7]:
1. Select a training pair from the training set (apply the first
input vector to the network input).
2. Calculate the output of the network.
3. Calculate the error between the network output and the
desired output (the target vector from the training pair)
4. Adjust the weights of the network in a way that minimiz-
Steps:

Fig. 10 Fingerprint Recognition ANN-PCA

es the error.
5. Repeat the steps 1 through 4 for each input vector in the
training set until the error for the entire set is acceptably low
[17].
5) ANN Testing stage
After the training stage has been completed, the testing and
validation stage are applied on different samples to check the performance of the network [17].
6) Matching stage
It is the final stage in fingerprint recognition system which is
used to identify the input fingerprint image. The algorithm of
matching process is to assign each fingerprint image (repre- sented by MDV vector), to one class named by Pi (where, i=101, 201...), for example finger print of the first person is cor-
1. Preprocessing: As mentioned before.
2. Feature extraction: As mentioned before.
3. Principal Component Analysis (PCA): After the minutiae have been extracted from fingerprint image, the minutiae matrix is created. PCA is used to reduce and compress the matrix data. Matrix data can be converted using PCA to principal component coefficients, principal component scores, or eigenvalues of the covariance matrix. The con- version of matrix data to principal component coefficients is developed and used here in fingerprint recognition sys- tem. Then these coefficients are reshaped to vectors form. The vectors of principal component coefficients are stored into the database and then they are treated as input of neural network.

IJSER © 2015 http://www.ijser.org

International Journal of Scientific & Engineering Research, Volume 6, Issue 5, May-2015 134

ISSN 2229-5518

4. Ann training: As mentioned before.
5. Ann testing: As mentioned before.
6. Matching: As mentioned before.

4 EXPERIMENTAL RESULTS

In the obtained experimental results we use sample of 100 fin- gerprint images with 500dpi resolution derived from FVC2002 database (http://bias.csr.unibo.it/fvc2002/ download.asp). FVC2002 database contains 800 fingerprint images captured with size of 504 × 480 pixels. These fingerprint images was captured with a live scanner. The fingerprint recognition sys- tem implementation was done using Matlab 7.10.0 (R2010a) installed on computer Intel Core I5 Processor, 2.27GHZ,
3.00GB RAM, and windows 7 Service Pack 1. The selected 100 fingerprint images have been provided to the fingerprint recognition system. In the preprocessing stage, as possible the fingerprint image is enhanced using FDCT, segmentation, normalization, orientation & frequency estimation, and Gabor filter. Then the thinning step is performed to remove redun- dant pixels. In post processing stage, the features are extracted from fingerprint image, and false minutiae are removed. The last stage is the matching stage, the fingerprints features are converted to vectors, training, validation, and testing the net- work. After extensive and powerful training the system simu- late the network to give an acceptable, robustness and proper recognition results. The characteristics of ANN used in the developed recognition system is given in table 2.

TABLE2 ANN CHARACTERISTICS

Fig. 11 Architecture of ANN using matlab.

In this experiment ANN network is trained through Feed- forward back propagation with scaled conjugate gradient al- gorithm. Fig. 12 displays the training progress of the network.

Fig. 12. ANN training result using matlab.

Training: It is used to adjust the network weights and biases according to its error (difference between actual network re- sult & desired result).
Validation: It is used to measure network generalization, and to stop training when generalization stops improving.
Testing: It is used to test the final solution in order to confirm the actual predictive power of the network.
The regression of training is shown in Fig. 13.

The ANN used in the experiment is shown in Fig. 11.

IJSER © 2015 http://www.ijser.org

International Journal of Scientific & Engineering Research, Volume 6, Issue 5, May-2015 135

ISSN 2229-5518

Fig. 13 Training regression result.

A part of the summary results (25-fingerprints images) of
ANN-MDV recognition system is shown in Table 3.

TABLE3 25 FINGERPRINTS SAMPLES RESULTS USING (ANN-MDV)

TABLE4 25 FINGERPRINTS SAMPLES RESULTS USING (ANN-PCA)

In the above tables the first column shows the serial number of samples, the second column is name of the sample, the third column represents the actual result given by the neural net- work simulation, and the fourth column shows the desired value of the output. The fifth column represents the recogni- tion state; the last column is the value of average recognition time.

5 COMPARISON & DISCUSSION

A comparative study between the 2-developed recognition systems is done based on system accuracy and ART. Also this study explains the performance, effectiveness, and powerful of each recognition system. Table 5 represents a comparison be- tween the 2 developed systems.

TABLE 5 COMPARISON BETWEEN ANN-MDV & ANN-PCA

Comparison parameter

ANN-MDV

ANN-PCA

Total number of fingerprints images

100

100

Number of recognized samples

91

98

Number of false recognized samples

9

2

Accuracy of the system

91%

98%

Average Recognition Time (ART)

0.2509

0.275

Also a part of the summary results (25 fingerprints) of finger- print recognition system using ANN-PCA is given in table 4.
The accuracy of the system can be calculated by (6).

𝑨𝑨𝑨𝑨𝑨𝑨𝑨𝑨 𝒐𝒐 𝒕𝒕𝒕 𝒔𝑨𝒔𝒕𝒕𝒔 = 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑟𝑒𝑐𝑜𝑔𝑛𝑖𝑧𝑒𝑑 𝑠𝑎𝑚𝑝𝑙𝑒𝑠 × 100 % (6)

𝑡𝑜𝑡𝑎𝑙 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑠𝑡𝑜𝑟𝑒𝑑 𝑠𝑎𝑚𝑝𝑙𝑒𝑠

Then the accuracy of ANN-MDV system is equal to:
91/100 × 100 = 91%.

IJSER © 2015 http://www.ijser.org

International Journal of Scientific & Engineering Research, Volume 6, Issue 5, May-2015 136

ISSN 2229-5518

Similarly the accuracy of ANN-PCA system is equal to:
98/100 × 100 = 98%.
Therefore ANN-PCA system is much better than ANN-MDV
system in terms of system accuracy.
Average recognition time can be calculated by (7).

[6 ] Jayanta Kumar Basu, Debnath Bhattacharyya, Tai-hoon Kim. "Use of Arti- ficial Neural Network in Pattern Recognition," International Journal of Software Engineering and Its Applications Vol. 4, No. 2, April 2010.

[7 ] Simon Haykin. "Neural Networks A comprehensive Foundation Second

Edition."

[8 ] R. Raja Sekar, K. Meena. "Fingerprint Recognition Using Multi-Resolution

𝑨𝑨𝒕𝑨𝑨𝑨𝒕 𝑨𝒕𝑨𝒐𝑨𝒓𝒓𝒕𝒓𝒐𝒓 𝒕𝒓𝒔𝒕 = 𝑠𝑢𝑚 𝑜𝑓 𝑟𝑒𝑐𝑜𝑔𝑛𝑖𝑡𝑖𝑜𝑛 𝑡𝑖𝑚𝑒 𝑜𝑓 𝑎𝑙𝑙 𝑠𝑎𝑚𝑝𝑙𝑒

𝑡𝑜𝑡𝑎𝑙 𝑛𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑠𝑡𝑜𝑟𝑒𝑑 𝑠𝑎𝑚𝑝𝑙𝑒𝑠

(7)

Techniques," International Journal of Electronics Communication and

Computer Engineering Volume 4, Issue (2) ICEA-2013, ISSN 2249–071X.

[9 ] Andreas Schmitt, Birgit Wessel, Achim Roth. "CURVELET APPROACH

For ANN-MDV system: the recognition time is calculated for
each fingerprint image, then the sum of these recognition
times of 100 fingerprints images is estimated in order to com-
pute ART which is equal to 0.2509.
In the same manner ART of ANN-PCA recognition system is
equal to 0.275. Hence ANN-MDV system is slightly better than
ANN-PCA in terms of ART.

6 SUMMARY, CONCLUSION AND FUTURE WORK

There are several ways of fingerprint recognition methods used to identify the person; we had developed two new effec- tive fingerprint recognition systems ("ANN-MDV" & "ANN- PCA"). The traditional methods can't provide satisfactorily results in case of unavailability of some fingerprint's features. As mentioned above one of the major advantage of neural network (which is the core function of the two developed methods) its capability of predicting and identifying the per- son fingerprint when some features are not found. The previ- ous explanation of PCA algorithm had stressed the fact that the main advantage of PCA is to compress data by reducing the matrix dimensions without losing much information, therefore reducing the size of fingerprint data base and results can be processed quickly. After discussion of the MDV algo- rithm we conclude that orientation direction of fingerprint image doesn't affect the performance of fingerprint recogni- tion system. From the previous comparison table we conclude the following:
1. The system accuracy of the ANN-PCA is much
higher than ANN-MDV.
2. ART of ANN- MDV is shorter than ANN-PCA.
Finally we concluded that ANN-PCA system is better than ANN-MDV system. Future work: we will use support vector machine SVM instead of neural network in order to improve fingerprint classification process. We can use circular Gabor filter with fast discrete curvelet transform FDCT to increase the efficiency of fingerprint image enhancement stage.

REFERENCES

[1 ] Handbook of Fingerprint Recognition Second Edition. D. Davide Maltoni, D. Dario Maio, Anil K. Jain, and Salil Prabhakar. New York Springer- Verlag, 2009.

[2 ] M.Tech Research Scholar, Vikram Panchal. "A Review on Finger Print

Recognition Systems". IJETCAS 13-183; © 2013.

[3 ] CHING-TANG HSIEH, SHYS-RONG SHYU. "Principal Component

Analysis for Minutiae Verification on Fingerprint Image," Proceedings of

the 7th WSEAS International Conference on Multimedia Systems & Signal

Processing, Hangzhou, China, April 15-17, 2007.

[4 ] Lindsay I Smith. "A tutorial on Principal Components Analysis," February

26, 2002.

[5 ] Divyakant T. Meva, C. K. Kumbharana, Amit D. Kothari. "The Study of

Adoption of Neural Network Approach in Fingerprint Recognition," Inter-

national Journal of Computer Applications (0975 – 8887) Volume 40– No.11, February 2012.

FOR SAR IMAGE DENOISING, STRUCTURE ENHANCEMENT, AND CHANGE DETECTION," In: Stilla U, Rottensteiner F, Paparoditis N (Eds) CMRT09. IAPRS, Vol. XXXVIII, Part 3/W4 --- Paris, France, 3-4 Septem- ber, 2009.

[10 ] L. Ravi Kumar, S. Sai Kumar, J. Rajendra Prasad, B. V. Subba Rao, P. Ravi

Prakash. "Fingerprint Minutia Match Using Bifurcation Technique," S Sai Kumar et al , International Journal of Computer Science & Communication Networks, Vol 2(4), 478-486, 2012.

[11 ] Avinash Pokhriyal, Praveen Sengar, Sushma Lehri. "Fingerprint Enhance- ment using Improved Orientation and Circular Gabor Filter," International Journal of Computer Applications (0975 – 8887) Volume 43– No.2, April

2012.

[12 ] Jyoti Rajharia, Dr. PC Gupta, Arvind Sharma. "Fingerprint-Based Identifi- cation System:–A Survey," International Journal of Computer Technology

and Electronics Engineering (IJCTEE) Volume 1, Issue 3, 2012.

[13 ] Ekta Thirani. "A Comparative Analysis of Fingerprint Enhancement Tech- nique through Absolute Mean Brightness Error and Entropy," IJRITCC |

JAN 2013, Available @ http://www.ijritcc.org.

[14 ] Nisha Negi, Sunil Semwal. "Minutiae Extraction and Pruning Based Fin- gerprint Identification with Pattern Classification by Radial Basis Function,"

International Journal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 , www.ijera.com ,Vol. 3, Issue 4, Jul-Aug 2013, pp.707-

712

[15 ] Neha Jain, S.K.Sharma,Ph.D, B.L.Pal. "Development of a Multimodal Biometric Identification and Verification System using Two Fingerprints," International Journal of Computer Applications (0975 – 8887) Volume 74– No.12, July 2013.

[16 ] Sudha S. Ponnarasi, M. Rajaram "Impact of Algorithms for the Extraction of

Minutiae Points in Fingerprint Biometrics" Journal of Computer Science 8 (9): 1467-1472, 2012 ISSN 1549-3636.

[17 ] Jyoti Rajharia, P.C. Gupta, PhD. "A New and Effective Approach for Fin- gerprint Recognition by using Feed Forward Back Propagation Neural Net- work," International Journal of Computer Applications (0975 – 8887) Vol-

ume 52– No.10, August 2012.

AUTHORS

1- Eng. Hany Hashem Ahmed is current- ly pursuing masters degree program in Department of Computer Science and Engineering, Faculty of Electronic Engineering, Menoufia University, Egypt. E-mail: hhahmd@hotmail.com Phone: +201097100936.

2- Prof. Hamdy M. Kelash is Emeritus Professor in the Department of Computer Science and Engi- neering, Faculty of Electronic Engineering, Menou- fia University. E-mail: HMK3947@yahoo.com.

3- Dr. Maha S. Tolba is lecturer in the Department of Computer Science and Engineering, Faculty of Electronic Engineering, Menoufia University. E- mail: maha_saad_tolba@yahoo.com.

4- Dr. Mohammed Badawy is lecturer in the Depart- ment of Computer Science and Engineering, Facul- ty of Electronic Engineering, Menoufia University. E-mail: MBMBadawy@yahoo.com.

IJSER © 2015 http://www.ijser.org