The study of human face recognition based curvelet transform and 2DPCA

As the wavelet transform cannot well represent curve singularity of human face images, this paper proposes an new algorithm based Curvelet and 2DPCA. For face images, we firstly perform the curvelet transform and get low frequency coefficients, which contain almost energy. Further we introduce 2DPCA with a exponential decay factor to reduce the dimensions and extract the feature vectors. Finally, the face recognition is realized according to the assembled matrix distance. Experimental results on ORL and Yale face database show that the proposed algorithm has high recognition rate and short recognition time. It is also robust to the change of pose, expression especially illumination.