Comparison of Face Recognition Approach through MPCA Plus LDA and MPCA Plus LPP

Document Sample
Comparison of Face Recognition Approach through MPCA Plus LDA and MPCA Plus LPP Powered By Docstoc
					                                                   (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                                             Vol. 8, No. 5, 2010




     Comparison of Face Recognition Approach
           through MPCA Plus LDA
              and MPCA Plus LPP
                                                       Shermina.J
                                            The Department of Computing,
                                         Muscat College, University of Stirling, UK
                                             Muscat, The Sultanate of Oman
                                                Shermina@mctcollege.com


                                                                    and configure of information[10]. The recognition of the
Abstract - Face recognition technology has been evolved as          face from a still or video clip can be obtained by
an enchanting solution to perform identification and the
                                                                    identifying or verifying one or more persons in the scene
verification of identity claims. By advancing the feature
extraction methods and dimensionality reduction techniques          using a stored database of faces. The solution includes the
in the application of pattern recognition, a number of face         process of segmentation of faces (face detection) from
recognition systems has been developed with distinct                cluttered scenes, feature extraction from the face
degrees of success. In this paper, we have presented a              regions, recognition or verification. In identification
biometric face recognition approach based on Multilinear            problems, the system reports the identity of an unknown
principal component analysis (MPCA) and Locality                    face from a database of known individuals. In verification
Preserving Projection (LPP) and the comparison of this to           problems, the system confirm or reject the claimed
the existing Multilinear principal component analysis               identity of the input face[12]. The organization of the
(MPCA) and Linear Discriminant Analysis (LDA). This
                                                                    paper is as follows: Review of related researches is
approach consists of four major steps: 1) Face image
preprocessing, 2) Dimensionality reduction using MPCA 3)            presented in Section 2. Description about the face
Feature Extraction using LPP and 4) Face recognition using          recognition methods used in our research is presented in
L2 similarity distance measure. Validation of this approach         section 3. Methodology of the approach is presented in
and its comparison with MPCA plus LDA is done with                  Section 4. Comparison of the two approaches is given in
FERET and AT&T face databases. Experimental results                 Section 5. The experimental results is bestowed in section
show the effectiveness of MPCA plus LPP in comparison               6 and finally the conclusions are summed up in Section 7.
with MPCA plus LDA in performance.
                                                                              II. REVIEW OF RELATED WORKS
    Keywords- Image Processing; Object Recognition; Face            Many face recognition methods have been developed in
Recognition; Image Compression; Multilinear Systems.                the past few decades. Most common feature extraction
                                                                    methods are principal component analysis (PCA) [1], [6]
                    I. INTRODUCTION                                 and linear discriminant analysis (LDA) [2], [5], [7], [8].
Biometrics is an automated method of recognizing a                  Another linear technique which is used for face
person based on their physical or behavioral                        recognition is Locality Preserving Projections (LPP) [3],
characteristics. Various types of recognition are available         [4], which finds an embedding that preserves local
commercially nowadays. Face recognition is the popular              information, and gains a face subspace that best detects
area of research in computer vision and the most                    the essential face manifold structure. PCA is a linear
successful applications of image analysis. It is the                method for unsupervised dimensionality reduction. [26].
biometric identification of a human’s face and matching             Kirby and Sirovich used PCA in 1987 in order to obtain a
it against a library of known faces. There are two                  reduced image[27]. Turk, Pentland, Moghaddam and
predominant approaches to face recognition, feature                 Starner extended the idea to eigenspace projections in the
based and appearance-based. The geometric feature-                  solution of face recognition[6]. PCA on tensor objects
based approach uses properties of facial features such as           requires their reshaping into vectors in a very high-
eyes, nose, mouth, chin and their relations for face                dimensional space, which not only results in high
recognition descriptors. The appearance-based face                  computational and memory demands but also breaks the
recognition approach operates directly on image based               natural structure and correlation in the original data.
representation. The whole face region is the raw input to           Useful representations can be obtained from the original
a recognition system and the facial features are processed          form and PCA extensions operating directly on the tensor
holistically involving an interdependency between feature           objects rather than their vectorized versions are emerging




                                                              198                               http://sites.google.com/site/ijcsis/
                                                                                                ISSN 1947-5500
                                                   (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                                             Vol. 8, No. 5, 2010

recently [11]. Mittal, N. and Walia, E. [17] introduced a           feature extraction methods on distinctly separate
fast PCA based face recognition algorithm. The                      subspaces seems to be more effective in the performance
efficiency of the proposed algorithm is verified using              of face recognition.
Indian face database, and the gained results prove an
improvement in performance of the proposed algorithm                              III. REOGNITION METHODS
as compared to the same with PCA method. Hossein                    Image based face recognition techniques can generally be
Sahoolizadeh et al. [18] have presented a face recognition          divided into two groups: 1) appearance-based that uses a
method on the basis of principal Component Analysis,                holistic texture features which can be applied to either
Linear Discriminant Analysis and neural networks. The               whole-face or particular regions in a face image, and 2)
proposed method is investigated on Yale face database.              feature based, that uses geometric facial features like
Experimental results on this database demonstrated the              mouth, eyes, brows, cheeks etc. and the geometric
effectiveness of the proposed method for face recognition           relationships existing between them [15]. Few
with a decrease number of misclassification while                   appearance-based methods are Principal Component
comparing with earlier methods. A two-dimensional                   Analysis (PCA), Independent Component Analysis
PCA (2DPCA) is proposed in [28] that constructs an                  (ICA), Linear Discriminant Analysis (LDA) and Locality
image covariance matrix using image matrices as inputs.             Preserving Projections (LPP). Among these we have
However, linear transformation is applied only to the               utilized the appearance-based recognition methods such
right side of image matrices so the image data is                   as MPCA, LDA and LPP.
projected in one mode only, resulting in poor
dimensionality reduction. A more general algorithm                  A. MULTILINEAR PRINCIPAL COMPONENT
named generalized low rank approximation of matrices                ANALYSIS (MPCA)
(GLRAM) was introduced in [31], which applies two                   Principal Component Analysis (PCA) is the widely used
linear transforms to both the left and right sides of input         technique in face recognition which is to reduce the
image matrices and results in a better dimensionality               dimensionality of a image set consisting of a large
reduction than 2DPCA. GLRAM is developed from the                   number of interrelated variables, by retaining most of the
perspective of approximation while the generalized PCA              variations in the image set. This is accomplished by
(GPCA) is proposed in [29] from the view of variation               transforming to a new set of variables, called the
maximization, as an extension of PCA. Later, the                    principal components which are uncorrelated but ordered
concurrent subspaces analysis (CSA) is formulated in                so that the first few retain most of the variation present in
[30] for optimal reconstruction of general tensor objects,          all of the original variables. This method of linear algebra
which can be considered as a generalization of GLRAM,               addresses single-factor variations in facial images. If
and the multilinear PCA (MPCA) introduced in [11]                   factors like lighting, viewpoint, expression, are also
targets at variation maximization for general tensor                permitted to modify facial images, eigenfaces undergoes
objects in the extension of PCA to the multilinear case,            difficulty. Multilinear Principal Component Analysis
which can be considered as a further generalization of              (MPCA) is the extension of PCA that employs
GPCA[26]. Wangmeng Zuo et al. [32] have depicted two                multilinear algebra and proficient of learning the
LDA-based methods, post-processed Fisherfaces and bi-               interactions of the multiple factors[11].
directional PCA plus LDA (BDPCA+LDA). pFisherfaces
utilizes 2D-Gaussian filter to smooth classical                     B. LINEAR DISCRIMINANT ANALYSIS (LDA)
Fisherfaces, and BDPCA+LDA is a LDA accomplished                    Linear Discriminant Analysis (LDA) is a classical
in the BDPCA subspace. Then they introduced a                       supervised linear subspace learning method for feature
combination framework where they merged these two                   extraction and dimension reduction that has been very
LDA-based approaches. Two popular face databases, the               successful and applied in various applications. It aims to
ORL and the FERET, are being used to compute the                    derive the most discriminative features and produces a
efficiency of the combination framework. The outcome                class specific feature space based on the maximization of
of the experiments revealed that the combination                    Fisher’s Discriminant Criterion, which is defined as the
framework      is     superordinate    to     pFisherfaces,         ratio of betweenclass scatter to withinclass scatter.
BDPCA+LDA, and other appearance-based methods in                    Classical LDA projects the data onto a lower-dimensional
terms of recognition accuracy. Deng Cai et al. [19] have            vector space such that the ratio of the between-class
proposed an appearance based face recognition method,               distance to the within-class distance is maximized, thus
called orthogonal Laplacian face in which face data may             achieving maximum discrimination. The optimal
be generated by sampling a probability distribution that            transformation can be obtained by applying the eigen
has support on or near a sub-manifold of ambient space.             decomposition on the scatter matrices[7].
This algorithm is based on locality preserving
projection(LPP) algorithm. Experimental results on three            C. LOCALITY PRESERVING PROJECTIONS (LPP)
face databases clearly demonstrated the efficiency of the           Locality Preserving Projection (LPP) is one of the linear
proposed algorithm. A unique feature extraction                     approximations of the nonlinear Laplacian Eigenmap.
technique is not felicitous when the dimensionality of              The locality preserving quality of LPP is responsible to
face images attempts to reach its peek and therefore an             be of particular use in information retrieval applications.
analytical study made using the combination of two                  In order to retrieve audio, video, text documents under a




                                                              199                                http://sites.google.com/site/ijcsis/
                                                                                                 ISSN 1947-5500
                                                    (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                                              Vol. 8, No. 5, 2010

vector space model, then it ultimately needs to do a                 variations must be related by some way to provide a
nearest neighbor search in the low dimensional space.                valid answer. Feature extraction or dimensionality
Since LPP is intended to preserve only local structure, it           reduction is a methodology to transform a high-
is probable that a nearest neighbor search in the low                dimensional data set into a low-dimensional equivalent
dimensional space will produce similar results compared              representation that assumes to retain most of the
to that in the high dimensional space. This allows an                information regarding the underlying structure or the
indexing scheme to access quickly [3].                               original physical phenomenon [16]. The main tendency
                                                                     of using feature extraction is its representation of data in
                  IV. METHODOLOGY                                    a lower dimensional space that computes through a linear
In face recognition problems, gallery and the probe are              or non-linear transformation satisfying certain properties.
the two types of the data sets used. The gallery set                 MPCA is a multilinear subspace learning method that
contains the set of data samples of unknown identities               extracts features directly from multi-dimensional objects.
and is used for training. The probe set is the testing set           MPCA receives the set of face image samples of the same
where the data samples of unknown identity are to be                 dimensions as input for feature extraction. The resultant
identified and classified by matching with corresponding             output of the MPCA is the dimensionally reduced feature
entries in the gallery set. MPCA accepts tensor samples              projection matrix of face images. Locality Preserving
of the dimension for feature extraction[11]. However, in             Projection (LPP) is one of the linear approximation
practice tensor object samples are of different                      obtained from the nonlinear Laplacian Eigenmap [9]. The
dimensions. Therefore, the input tensors need to be                  dimension reduced feature projection matrices of face
normalized to standard dimensions and then                           image samples obtained using MPCA is then fed as an
preprocessing procedure is used to remove the noise, if              input to the LPP algorithm. The dimensional reduced
required. The normalized tensor samples are then                     feature matrices of the training sample images obtained
centered by subtracting the mean obtained from the                   using the MPCA and LPP techniques are stored in a
gallery tensors. From the gallery set, a set of eigentensors         database. While we are testing the face images, the
is obtained with reduced dimensionality and each entry of            aforesaid techniques are applied to generate the feature
a projected tensor feature can be viewed as a scalar                 matrix and thereby a similarity measure is carried out on
feature corresponding to a particular eigentensor. Some              the sample face images. Various face recognition systems
of the small variation and noise are removed in the                  may use different distance measures while matching
projection. MPCA is an unsupervised technique without                query images with the nearest database images. Our Face
taking class labels into account. Hence, variation captured          recognition approach used here is performed using L2
in the projected tensor subspace includes both the within-           distance measure. The L2 distance is computed between
class variation and the between-class variation. A larger            the face images present in the database and the query
between-class variation to the within-class variation                image for matching process. The similarity distance
indicates good class separability. Hence the feature                 measure for a pair of face images is computed in which a
selection strategy selects eigentensors according to class           threshold determines whether the face pair is classified as
discrimination power which is the ratio of between-class             same or different.
variation over within-class variation. In classification[2],
the distance between feature vectors is of importance as it          B. MPCA PLUS LDA APPROACH
determines the performance of the classification module.             The use of Linear Discriminant Analysis to perform
Distance between unknown probe image and gallery                     feature extraction combined with the projected
images stored in database are measured by any of the                 multilinear arrays produces the approach MPCA+LDA.
distance measures L1, L2, angle or             Mahalanobis           The feature vector can be used directly for recognition,
distance.                                                            and a classical linear discriminant analysis (LDA) can
                                                                     also be applied to obtain an MPCA+LDA approach for
A. MPCA PLUS LPP APPROACH                                            recognition [23], similar to the PCA+LDA approach.
Image processing techniques such as normalization and                MPCA+LDA extend PCA to use the entire face image as
resizing of the face images            are employed in               a tensor object inorder to preserve the relationship
preprocessing in order to improve the face image. Next,              between neighboring pixels. LDA make use of class
the feature extraction is achieved by merging the MPCA               information in data and attempts to minimize the within
along with LPP to calculate the feature projection                   class scatter and to maximize the between class scatter.
matrices. The feature projection matrices project the face           When using the AR database, the recognition rate is
images to lower dimension and generate the face feature              proved to be better for full image face.
vectors. While testing, the test face sample is mapped
onto a feature matrix with the assistance of MPCA and                   V. COMPARISON OF MPCA PLUS LDA WITH
LPP. The face recognition can be done by comparing the                               MPCA PLUS LPP
test feature matrix with the enrolled face features in the           The proposed approaches of MPCA+LDA and
database using L2 distance. When processing a face, the              MPCA+LPP is tested using the FERET database [13] and
features like variations in light, image quality, persons’           AT&T database [21] of faces. Performance is measured
pose, facial expressions and more should be taken into               by procedures applied to FERET facial images and
account. To identify correct individuals successfully these          AT&T facial images. In particular, all the images were




                                                               200                                http://sites.google.com/site/ijcsis/
                                                                                                  ISSN 1947-5500
                                                               (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                                                         Vol. 8, No. 5, 2010

preprocessed using normalization and then by resizing of                      techniques for recognizing the facial images. In this
the images. The images are divided into training set and                      article, the comparison of Face recognition approaches
test set. The training set is used to initialize and prepare                  MPCA plus LDA and MPCA plus LPP is presented. The
the system to recognize the arbitrary images. The test set                    combined appearance based technique such as MPCA
is the set of images used to evaluate the performance of                      and LPP yield to produce a high face recognition rate
the system once the training phase is completed. Here, in                     compared to MPCA and LDA technique. Experimental
FERET database, we use nearly 80 images for training                          results on FERET and AT&T database demonstrated the
and 160 images for testing. In AT&T database, we use                          effectiveness of the proposed approach with improved
100 images for training and 200 images for testing                            recognition accuracy. In future, fusion of MPCA plus
process. The experiments were conducted using the L2                          other face recognition approaches could be experimented
distance measure. The common way to measure the                               and the comparitive analysis could be done to verify the
biometric recognition accuracy is to compute the false                        performance of the approach.
acceptance rate (FAR) and false rejection rate (FRR).
FAR is the percentage of incorrect acceptances. FRR is                                         ACKNOWLEDGMENT
the percentage of incorrect rejections. The accuracy                          I would like to thank Dr.V.Vasudevan for providing the
measurement of the overall approach is computed as 100-                       guidance     in the    preliminary discussions on the
(FAR/FRR)/2. Genuine acceptance rate (GAR) is an                              comparison of the two approaches used in order to
overall accuracy measurement of the approach.                                 proceed further in my area of research.

          VI. EXPERIMENTAL RESULTS                                                                      REFERENCES
The percentage of accuracy is given in the following                          [1] H. Moon, P.J. Phillips, “Computational and Performance aspects of
                                                                              PCA based Face Recognition Algorithms”, Perception, vol. 30, pp. 303-
table 1.                                                                      321, 2001.
                                                                              [2] P.N. Belhumeur, J.P. Hepanha, and D.J. Kriegman, “Eigen faces
        Table 1: Accuracy rates on FERET and AT&T databases                   vs.fisherfaces: recognition using class specific linear projection,”IEEE.
                                                                              Trans. Pattern Analysis and Machine Intelligence, vol. 19, no. 7, pp.
                                Face Recognition Approach                     711-720, 1997.
        Database
                                                                              [3] X. He and P. Niyogi, “Locality preserving projections, ”in Advances
                             MPCA+LDA                 MPCA+LPP                in Neural Information Processing Systems. Cambridge, MA: MIT Press,
                                                                              2003.
        FERET                  93.80%                   96.50%                [4] Li, Ruidong; Zhu, Lei; Zou, Shaofang; “Face Recognition Based on
                                                                              an Alternative Formulation of Orthogonal LPP”, Control and
                                                                              Automation, 2007, IEEE International Conference on May 30 2007-
         AT&T                   95%                     96.50%
                                                                              June 1, pp.2374 – 2376, 2007.
                                                                              [5] P. N. Belhumeur, J. P. Hespanha and D. J. Kriegman, “Eigenfaces
                                                                              vs. Fisherfaces: Recognition Using Class Specific Linear Projection”,
The comparative results are analyzed with the recognition                     IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 19, No. 7,
rates of the feature extraction methods such as MPCA                          pp. 711-720, 1997.
plus LDA and MPCA plus LPP by means of charts. The                            [6] M. Turk and Pentland, “Face Recognition Using Eigenfaces”, in
charts clearly shows that the recognition performance of                      Proc. IEEE International Conference on Computer Vision and Pattern
                                                                              Recognition, Maui, Hawaii, 1991.
the approach MPCA plus LPP is more efficient when                             [7] J. Yang, Y. Yu, and W. Kunz, “An Efficient LDA Algorithm for
compared to MPCA plus LDA. The face recognition                               Face Recognition”, The sixth International Conference on Control,
results of MPCA+LPP is compared with MPCA+LDA                                 Automation, Robotics and Vision, Singapore, 2000.
based on the measures such as false acceptance rate, false                    [8] W. Zhao, R. Chellappa, and P. J. Phillips, “Subspace Linear
                                                                              Discriminant Analysis for Face Recognition”, Technical Report CAR-
rejection rate, genuine acceptance rate and recognition                       TR-914, Center for Automation Research, University of Maryland,
accuracy. The comparative result of each recognition rate                     1999.
is shown in the following chart.                                              [9] M. Belkin and P. Niyogi, “Laplacian Eigenmaps and Spectral
                                                                              Techniques for Embedding and Clustering,” Advances in Neural
  100
                                                                              Information Processing Systems 14, Vancouver, British Columbia,
                                                                              Canada, 2002.
   90
                                                                               [10] Borut Batagelj and Franc Solina, “ Face Recognition in Different
   80                                                                         Subspaces: A Comparative Study”
                                                       FERET Database
   70                                                  MPCA+LDA               [11] Haiping Lu, K.N. Plataniotis, A.N. Venetsanopoulos, "MPCA:
   60                                                                         Multilinear Principal Component Analysis of Tensor Objects", IEEE
                                                       FERET Database
 % 50                                                  MPCA+LPP               Transactions on Neural Networks, Vol. 19, No. 1, January 2008.
                                                                              [12] Zhao, W. Chellappa, R., Phillips, P. J. and Rosenfeld, A., “Face
   40                                                  AT&T Database
                                                       MPCA+LDA               Recognition: A Literature Survey”, ACM Computing Survey,
   30                                                                         December, pp: 399-458, 2003.
                                                       AT&T Database
   20                                                  MPCA+LPP               [13] P.J. Phillips, H. Moon, P. Rizvi, and P. Rauss, ˝The FERET
   10                                                                         evaluation method for face recognition algorithms, ˝ IEEE Transactions
    0
                                                                              on Pattern Analysis and Machine Learning, vol. 22, pp. 1090-1104,
         FAR (%)   FRR (%)   GAR (%)   Accuracy (%)                           2000.
                                                                              [14] S. T. Gandhe, K. T. Talele, and A.G.Keskar, "Face Recognition
                                                                              Using Contour Matching", IAENG International Journal of Computer
                                                                              Science, Vol: 35, No: 2, 2008.
                 VII. CONCLUSION                                              [15] M. H. C. Law and A. K. Jain, “Incremental nonlinear
Face Recognition is the active area of research over the                      dimensionality reduction by manifold learning,” IEEE Trans. Pattern
past two decades which have resulted in successful                            Anal. Mach. Intell., Vol. 28, No. 3, pp. 377–391, Mar. 2006.




                                                                        201                                      http://sites.google.com/site/ijcsis/
                                                                                                                 ISSN 1947-5500
                                                                  (IJCSIS) International Journal of Computer Science and Information Security,
                                                                                                                            Vol. 8, No. 5, 2010

[16] V.V. Starovoitov, D.I Samal, D.V. Briliuk, "Image Enhancement
for Face Recognition", International Conference on Iconics,
St.Petersburg, Russia, 2003.
[17] Mittal, N. Walia, E., "Face Recognition Using Improved Fast PCA
Algorithm", IEEE Congress on Image and Signal Processing, Vol: 1,
pp: 554 - 558, 2008.
[18] A. Hossein Sahoolizadeh, B. Zargham Heidari, and C. Hamid
Dehghani, "A New Face Recognition Method using PCA, LDA and
Neural Network", International Journal of Computer Science and
Engineering, Vol: 2, No :4, 2008.
[19] Deng Cai, Xiaofei He, Jiawei Han, Hong-Jiang Zhang, "Orthogonal
Laplacianfaces for Face Recognition", IEEE Transactions On Image
Processing, 2006
[20] Ben Niua, Qiang Yang, Simon Chi Keung Shiu, Sankar Kumar Pal,
"Two-dimensional Laplacianfaces method for face recognition", Pattern
Recognition, Vol: 41, pp: 3237 – 3243, 2008.
[21] "AT&T database of faces" from http://www.cl.cam.ac.uk/
research/dtg/attarchive/ facedatabase.html
[22] M.A.O. Vasilescu, D. Terzopoulos. “Multilinear analysis for facial
image recognition,” In Proc. Int. Conf. on Pattern Recognition, Quebec
City, August 2002.
[23] J.Harguess, J.K.Aggarwal, “An average half face in 2D and 3D
Face recognition” in Pattern Recognition and Machine Vision, 2010
[24] Haiping Lu, K. N. Plataniotis and A. N. Venetsanopoulos,
"Boosting Discriminant Learners for Gait Recognition Using MPCA
Features", EURASIP Journal on Image and Video Processing, 2009,
doi:10.1155/2009/713183.
[25] Kela, N. Rattani, A. Gupta, P, "Illumination Invariant Elastic
Bunch Graph Matching for Efficient Face Recognition", Computer
Vision and Pattern Recognition Workshop, pp: 42, 2006.
[26] Haiping Lu, K.N.Plataniotis, A.N.Venetsanopoulos, “Uncorrelated
Multilinear Principal Component          Analysis through Successive
Variance Maximization” in 25th International Conference on Machine
Learning, 2008
[27] M.Kirby, L.Sirovich, “Application of the Karhunen-Loeve
Procedure for the characterization of Human Faces”, 1990
[28]  Yang, J., Zhang, D., Frangi, A. F., & Yang, J. (2004). Two-
dimensional PCA: a new approach to appearance-based face
representation and recognition. IEEE Trans.Pattern Anal. Mach. Intell.,
26, 131-137.
[29] Ye, J., Janardan, R., & Li, Q. (2004). GPCA: An efficient
dimension reduction scheme for image compression and retrieval. The
10th ACM SIGKDD Int. Conf. on Knowledge Discovery and Data
Mining (pp. 354{363).
[30] Xu, D., Yan, S., Zhang, L., Zhang, H.-J., Liu, Z., & Shum;, H.-Y.
(2005). Concurrent subspaces analysis. Proc. IEEE Computer Society
Conf. on Computer Vision and Pattern Recognition (pp. 203{208).
[31] Ye, J. (2005). Generalized low rank approximations of matrices.
Machine Learning, 61, 167-191.
[32] Wangmeng Zuo, Kuanquan Wang, David Zhang, Hongzhi Zhang,
"Combination of two novel LDA-based methods for face recognition",
Advanced Neurocomputing Theory and Methodology, Vol: 70, No: 4-6,
pp: 735-742 , January 2007.

                       AUTHORS PROFILE

                           Shermina.J received the Master degree in
                           Computer Science and Engineering in the year
                           2002 from Madurai Kamaraj University, India.
                           She began to pursue the Ph.D degree in the
                           Department of Information Technology in
                           Kalasalingam University, India in 2008.
                           Currently she is working as a lecturer in Muscat
                           College affiliated to the University of Stirling,
                           UK. She is a member of IEEE and the current
research interest is in the field of image processing.




                                                                               202                            http://sites.google.com/site/ijcsis/
                                                                                                              ISSN 1947-5500

				
DOCUMENT INFO
Description: IJCSIS invites authors to submit their original and unpublished work that communicates current research on information assurance and security regarding both the theoretical and methodological aspects, as well as various applications in solving real world information security problems.