Face recognition based on Gabor phase
【摘要】：正This paper proposes a novel representation method for face recognition. Given a face image, its Gabor phases instead of the traditional Gabor magnitudes are extracted by convolving the face image with multiple Gabor filters at different scales/frequencies and orientations. Then, linear discriminant analysis (LDA) is used to reduce the high dimension feature vector and extract the efficient feature for face recognition. In order to overcome the singularity of intra-class scatter matrix of LDA, principal component analysis (PCA) reduces the vector dimension before applying LDA. The experimental results on the FERET databases demonstrate the effectiveness of the proposed method.