Face recognition based on Second Generation of Curvelet Transform and Kernel Principal Component Analysis

Abstract

A new face recognition method is proposed by adopting the Second Generation of Curvelet Transform (SGCT) and Kernel Principal Component Analysis (KPCA). Based on KPCA, the face recognition algorithm can extract nonlinear image features and show better performance under the conditions of small sample training. However, the disadvantage of KPCA is the image information redundancy, which reduces the recognition performance. Traditional wavelet transform method of preprocessing removes irrelevant details of identification, but in high-dimensional image signal, the wavelet analysis is not the optimal method. In this paper, the new multi-scale geometric analysis, SGCT, is proposed to preprocess the image in order to reduce the high dimensional operators and improve accuracy of KPCA. Based on ORL database, experimental results show that the proposed method has a faster recognition speed and higher recognition accuracy than the traditional methods.