Applications of Block Linear Discriminant Analysis for Face Recognition

Face recognition is an important issue in image analysis and understanding. Linear discriminant analysis (LDA) has been widely used in face recognition. However, the LDA-based face recognition methods usually encountered the small sample size (SSS) problem. The SSS problem occurs when the number of samples is far smaller than the dimensionality of the sample space. Therefore, this paper proposed a modi_ed LDA (called block LDA) to divide the gradient image into several non-overlapping subimages of the same size, in order to increase the quantity of samples and reduce the dimensions of the sample space. In addition, to reduce the inuence of illumination variations, face images were transferred to gradient image. Experimental results show that the proposed method indeed solves the SSS problem with a good recognition rate.