Research On Linear Feature Extraction Methods In Face Recognition

Feature extraction is an important research branch of pattern recognition field; it is the key to solve the face recognition problems. Linear feature extraction methods,including Principal component analysis(or K-L transform)and Fisher linear discriminant analysis,is the classical and popular technique for feature extraction. In this paper, we carried trough some research to the current leading algorithms of linear feature extraction,and the proposed algorithms can be sure successfully applied on the face recognition.In face recognition,FLDA always encounter high dimensionality and small sample size problem,which usually leads to singularity of the within-class scatter matrix,which is a trouble for calculation of Fisher optimal discriminant vectors. Many researchers have proposed many feature extraction methods to solve the small sample size problem, but these methods only reflect the image changes between the rows or columns, and changes in the other direction are omitted, while changes in the other direction is also useful for face recognition. To solve the problem, a novel framework called Tri-direction 2D Fisher Discriminant Analysis(T2D-FDA) is proposed to deal with the Small Sample Size (SSS) problem in conventional One-Dimensional Linear Discriminant Analysis (1D-LDA). Moreover, the essence of D2D-FDA is investigated, and the equivalence of the left-multiplying 2DFDA of the original image matrices and the left-multiplying D2D-FDA of diagonal image matrices is verified if each column is viewed as a computational unit. Different from the 1D-LDA based approaches, 2D-FDA is based on 2D image matrices rather than column vectors, so the image matrix does not need to be transformed into a long vector before feature extraction. The advantage arising in this way is that the SSS problem does not exist anymore because within-class scatter matrices constructed in 2D-FDA is of full-rank. The proposed method is applied to face recognition where only few training images exist for each subject. Experiment results show T2D-FDA outperforms the current linear subspace methods in small sample size problem on two public databases: ORL and Yale face database.