Research of insensitive feature extraction under varying illumination conditions

Achieving illumination invariance in the insensitive feature of face images remains one of the most challenging aspects of automatic face recognition. In this paper, a novel mirror-like Gabor feature is proposed for its robust and efficient matching. First, the process of mirror-like Gabor feature extraction is presented. Second, it gives a theoretical analysis from the Fisher criterion and statistical correlation points of view. Finally, the comparison experiment between mirror-like Gabor feature and Gabor feature is performed on the YaleB image databases, and the experimental results indicate that the former has the powerful robustness.