Abstract:

It is currently a hot research topic that how to improve the robustness of the face recognition and reduce the influence of light,expression and occlusion. A robust face recognition method combining sparse representation and wavelet transform is proposed in this paper. Firstly, the original images are decomposed into four subband images（LL、LH、HL、HH）by applying wavelet transform，and the four different band images are chosen as basic functions to construct the dictionary.Secondly,the subband LL of testing images is decomposed sparsely. Finally, the classifier based minimal reconstruction error is used for identification. The experiment result on Yale face databases shows this method is better than the comparative method.

Key words: face recognition;wavelet transform;sparse representation;orthogonal matching pursuit;compressed sensing;robustness