Using the original and symmetrical face training samples to perform collaborative representation for face recognition

Abstract More training samples are able to reveal more possible variation of the illumination, expression and poses and are consequently beneficial for correct classification. However, in real-world applications, there are usually only a limited number of available training samples. Therefore, it is hard to effectively improve the accuracy of face recognition. The symmetry of face is of great importance to face recognition. In this paper, based on the symmetry of the face, the new mirror training samples are first generate new samples. Then the original training samples and the generated symmetry training samples are, respectively, used to perform collaborative representation based classification method. Finally, the scheme of the score level fusion is adopted to integrate the original training samples and symmetrical face training samples for ultimate face recognition by assigning a larger weight to the original training samples. The experimental results show that the proposed method can classify the face with a high accuracy.