Abstract: Face recognition is common biometric identification,widely used in access control,public security and justice,and so on.Existing face recognition methods are greatly influenced by the change of acquisition conditions on illumination,facial expression,pose and occlusion,limiting the application of face recognition technology.This paper proposed a face recognition method based on curvelet transform and cosine rules,to improve the robustness on acquisition conditions.First,we executed curvelet transform on face image,and detected key points according to curvelet coefficients.Then,we extracted curvelet features from key points with different scales and directions,and obtained a face descriptor.Finally,the optimal matching result was computed through cosine rules,cumulative and extreme operation.Experiments show that the proposed method is robust to changes of illumination,facial expression,pose and occlusion,and has good recognition performance.

Key words: Face recognition,Curvelet transform,Cosine rules,Scales adjustment