Face recognition based on improved Retinex and sparse representation

Abstract In this paper, we proposed a method based on improved Retinex theory and sparse representation to deal with the difficulties for face recognition under inhomogeneous illumination. In our work, the total variation model was introduced to optimize the parameters of Retinex and the illumination insensitive features were extracted as the dictionary of sparse representation. Finally, the facial images could be recognized by the proposed algorithm. The experimental results on different benchmark face databases indicated that the proposed approach could be more efficient than traditional methods for face images under uncontrolled illumination conditions.