A New Face Recognition Method Based on 2DWPCA

A new method based on Two-Dimensional Within-class Principal Component Analysis (2DWPCA) is proposed for face recognition in this paper. First, the within-class image scatter matrix of each class is calculated by using the training face samples in each class, respectively. Then, according to the within-class image scatter matrix of each class, the optimal eigenvectors of each class are computed, and are selected as the optimal projection axes of each class. Finally, a minimal distance classifier is employed to classify the given test samples. The proposed method is evaluated on the NUST603 face database. Experimental results demonstrate that the method proposed in this paper is effective and feasible.