(38-4) 43 * << * >> * Russian * English * Content * All Issues

On the solution of the image recognition problem by a principal component method and linear discriminant analysis
V.V. Mokeyev , S.V. Tomilov

 

South Ural State University (National Research University)

PDF, 1184 kB

Full text of article: Russian language.

DOI: 10.18287/0134-2452-2014-38-4-871-880

Pages: 871-880.

Abstract:
In the paper, some aspects of image analysis based on PCA (Principal Component Analysis) and LDA (Linear Discriminant Analysis) are considered. The main idea of this approach is that, firstly, we project the face image from the original vector space to a face subspace via PCA, secondly, we use LDA to obtain a linear classifier. In the paper, the efficiency of application of PCA and LDA to a problem of recognition of face images without their preliminary normalization is investigated. When the number of images in a class is not large, it is proposed that the training set is supplemented by images obtained by rotating, scaling and mirroring. In the images from the ORL and Feret databases, the influence of the training set expansion on the quality of recognition of unnormalized face images is studied. Also, a problem of increasing the efficiency of principal component calculation for large image samples is addressed. A linear condensation method is used as a new technique to calculate the principal components of a large matrix. The accuracy and performance of the developed algorithm are evaluated.

Key words:
face recognition, principal component analysis, linear discriminant analysis, linear condensation method.

Citation:
Mokeyev VV, Tomilov SV. On the solution of the image recognition problem by a principal component method and linear discriminant analysis. Computer Optics 2014; 38(4): 871-880. DOI: 10.18287/0134-2452-2014-38-4-871-880.

References:

  1. Kirby, M. Application of the KL procedure for the characterization of human faces / M. Kirby, L. Sirovich // IEEE Transactions on Pattern Analysis and Machine Intelligence. – 1990. – Vol. 12(1). – P. 103-108.
  2. Etemad, K. Discriminant Analysis for Recognition of Human Face Images / K. Etemad, R. Chellappa // Journal of the Optical Society of America A. – 1997. – Vol. 14(8). – P. 1724-1733.
  3. Lu, J. Face Recognition Using LDA-Based Algorithms / J. Lu, K.N. Plataniotis, A.N. Venelsanopoulos // IEEE Trans, on Neural Networks. 2003. – Vol. 14(1). – P. 195-200.
  4. Belhumeur, P.N. Eigenfaces vs Fisherfaces: recognition using class specific linear projection / P.N. Belhumeur, J.P. Hespanha, D.J. Kriegman // IEEE Transactions on Pattern Analysis and Machine Intelligence. – 1997. – Vol. 19. – P. 711-720.
  5. Martinez, А.М. РСА versus LDA / А.М. Marline, А.С. Kak // IEEE Trans, on Pattern Analysis and Machine Intelligence. – 2001. – Vol. 23(2). – P. 228-233.
  6. Kukharev, G.A. Algorithms of two-dimensional principal component analysis for face recognition / G.A. Kukharev, N.L. Shchegoleva // Computer Optics. – 2010. – Vol. 34(4). – P. 545-551. (In Russian).
  7. Mokeev, А.V. On accuracy and performance principal component synthesis method / A.V. Mokeev // Business Informatics. – 2010. – Vol. 3(18). – P. 64-68. – (In Russian).
  8. Mokeyev, V.V. On effectiveness increase of principal components computation in image analysis problem / V.V. Mokeyev // Digital Signal Processing. – 2011. – Vol. 4. – P. 29-36. – (In Russian).
  9. Grinenko, N.I. Problems of studying vibrations of structures by the finite-element method / N.I. Grinenko, V.V. Mokeev // International Applied Mechanics. – 1985. – Vol. 21(3). – P. 25-30. – (In Russian).
  10. Mokeev, V.V. On the problem of finding the eigenvalues and eigenvectors of large matrix system, arising in use a finite element method / V.V. Mokeev // Computational Mathematics and Mathematical Physics. – 1992. – Vol. 32(10). – P. 1652-1657. – (In Russian).
  11. Mokeyev, V.V. On solution of small sample size problem with linear discriminant analysis in face recognition / V.V. Mokeyev, S.V. Tomilov // Business Informatics. – 2013. – Vol. 1(18). – P. 37-43. – (In Russian).
  12. Phillips, P.J. The FERET evaluation methodology for face recognition algorithms / P.J. Phillips, H. Moon, P.J. Rauss, S. Rizvi // IEEE Transactions on Pattern Analysis and Machine Intelligence. – 2000. – Vol. 22(10). – P. 1090-1104.
  13. Gu, X. Regularized locality preserving discriminant analysis for face recognition / X. Gu, W. Gong, L. Yang // Neurocomputing. – 2011. – Vol. 74. – P. 3036-3042
  14. Zhang, T. Patch alignment for dimensionality reduction / T. Zhang, D. Tao, X. Li [et al.] // IEEE Transaction on Knowled­ge and Data Engineering. – 2009. – Vol. 21(9). – P. 1299-1313.
  15. Yang, L. Null space discriminative locality preserving projections for space recognition / T. Zhang, G.X. Gu [et al.] // Neurocomputing. – 2008. – Vol. 71. – P. 2045-2054.
  16. Pang, Y.H. Regularized locality preserving discriminant embedding for face recognition / Y.H. Pang, A.B.J. Teoh, F.S. Abas // Neurocomputing. – 2012. – Vol. 77. – P. 156-166.
  17. Li, S.Z. Non linear mapping from multi-view face patterns to a gaussian distribution in a low dimensional space / S.Z. Li, X. Rong, L.Z. Yu, Zhang HongJiang // in: Proceedings of the IEEE ICCV Workshop on Recognition, Analysis, and Tracking of Face sand Gestures in Real-Time Systems, 2001. – P. 47-54.
  18. Yan, S.C. Graph embedding and extensions: a general framework for dimensionality reduction / S.C. Yan, Dong Xu, Benyu Zhang, Hong-Jiang Zhang, Qiang Yang, S. Lin // IEEE Transactions on Pattern Analysis and Machine Intelligence – 2007. –Vol. 29(1). – P. 40-51.
  19. Cai, D. Locality sensitive discriminant analysis / D. Cai, X. He, K. Zhou, J. Han, H. Bao // in: Proceedings of the International Joint Conference Artificial Intelligence, 2007. – P. 708-713.
  20. Parlett, B.N. The symmetric eigenvalue problem / B.N. Par­lett. – New Jersey, Englewood Cliffs: Prentice-Hall, 1980.
© 2009, IPSI RAS
151, Molodogvardeiskaya str., Samara, 443001, Russia; E-mail:journal@computeroptics.ru; Tel: +7 (846) 242-41-24 (Executive secretary), +7 (846) 332-56-22 (Issuing editor), Fax: +7 (846) 332-56-20