Face Recognition Based on Two-Dimension Kernel Principal Component Analysis and Fuzzy Maximum Scatter Difference

Considering the "nonlinear","outer classes" and "hard classifier" problem in two-direction maximum scatter difference discriminant analysis method, a new method(2DKFMSD) of face recognition based on two-dimension kernel principal component(K2DPCA) and fuzzy maximum scatter difference (FMSD) is developed in this paper. Firstly, the K2DPCA overcomes the limitations of the traditional PCA method and can extract the nonlinear structures features in faces efficiently. Secondly, selecting the eigenvectors that between-class scatter is greater than within-class scatter after projection as optimal projection axis. and the distribution information of samples is represented with fuzzy membership degree in the FMSD. Finally, it uses the nearest neighbor classifier for face recognition. The experiment results on ORL and YALE face databases show that the 2DKFMSD is better than other methods.