Illumination normalization based on 2D Gaussian illumination model

Achieving illumination invariance in the presence of varying lighting conditions remains one of the most challenging aspects of automatic face recognition. In this paper, a novel approach for illumination normalization under varying lighting conditions is presented. This method is based on a 2D Gaussian illumination model, which is first proposed in this paper. This model can be used for contrast stretching in the “dark” areas on the face images. In our method, we choose Quadtree to o locate the shadows, and then apply the 2D Gaussian illumination model to adjust contrast of these dark areas, last utilize the symmetrical property of human face to obtain the illumination invariance features of the face images. The proposed algorithm has been evaluated based on the Yale B database. The experimental results show that our algorithms can significantly improve the performance of face recognition under uneven illumination conditions.