Abstract:
This paper first analyzes the Principal Component Analysis (PCA),2-Dimensional PCA (2DPCA) and Complete 2D PCA (C2DPCA) shortcomings,for the PCA method can not solve the problem of small sample,2DPCA and C2DPCA treat equally to all identifying information,To solve the problem,this paper propose a locality preserving C2DPCA method.Firstly,the face image is divided into five regions,To preserve eyes,lips and contour.reduce the degree of scatter to other region,then use C2DPCA method to dimensionality reduction and feature extraction,it shown that this method reduce the dimension of feature matrix and improve recognition rate through experiment by ORL face database compared with C2DPCA.and the recognition rate is superior to classic Fisherfaces and ICA methods.