Efficient human face recognition based on shape and texture

Shape and texture of a human face are useful and complementary´ information for face recognition. Hausdorff distance has been successfully applied for face recognition, which can measure the difference between two face images based on their respective edge representations. However, this distance measure does not consider the texture information about human faces. In this paper, a new Hausdorff distance that combines both the shape and texture information is proposed for human face recognition. The texture information is represented by means of the Gabor wavelets. This new method has the advantages of invariant to illumination and low memory requirement as Hausdorff distance and a high recognition performance as the Gabor wavelets. Experiments based on the Yale database. AR database and ORL database show that our proposed algorithm can achieve recognition rates of 85.3%. 95.9%. and 85.7%. respectively.