Mirror-Like Gabor Features for Face Recognition under Varying Illumination Conditions

As is well known, face recognition under varying illumination conditions remains one of the most challenging tasks. How to achieve illumination invariant features is the key issue of face recognition. This paper provides a novel face recognition method using mirror-like Gabor features (MGF). The paper first presents the process of even/odd MGF extraction. It then demonstrates that the even MGF are superior to the odd ones in terms of robustness and efficient matching, providing relevant theoretical analysis from the point of view of the Fisher criterion and statistics. Finally, it describes comparison experiments on the YaleB face database and offers valuable conclusions.