Kernel maximum scatter difference based feature extraction and its application to face recognition

This paper formulates maximum scatter difference (MSD) criterion in the kernel-including feature space and develops a two-phase kernel maximum scatter difference criterion: KPCA plus MSD. The proposed method first maps the input data into a potentially much higher dimensional feature space by virtue of nonlinear kernel trick, and in such a way, the problem of feature extraction in the nonlinear space is overcome. Then the scatter difference between between-class and within-class as discriminant criterion is defined on the basis of the above computation; therefore, the singularity problem of the within-class scatter matrix due to small sample size problem occurred in classical Fisher discriminant analysis is avoided. The results of experiments conducted on a subset of FERET database, Yale database indicate the effectiveness of the proposed method.