Illumination Invariant Face Recognition by Local Image Descriptor in Logarithm Domain

Abstract

Making recognition more reliable under uncontrolled lighting conditions is one of great challenges for practical face recognition systems. In this paper, inspired by the local binary pattern, we present a simple but effective illumination invariant image descriptor in the logarithm domain. In the proposed approach, the logarithm transform is used firstly. Then the subtraction of each pixel with the neighborhood is implemented. Based on the reflectance model [2], and the assuming that the illumination is constant if the facet is small enough, the differentials are illumination invariant and can be treated as the illumination invariant features. In order to make the recognition task simpler, the sum of the differentials of each pixel with the neighborhood is considered as each pixel's new vale and we use the new image to recognize directly. The experimental results on the Yale B and Extended Yale B prove the out performance of the proposed method compared to other existing methods.