Face Recognition Using Spatially Smooth and Maximum Minimum Value of Manifold Preserving

Subspace face recognition methods have attracted considerable interests in recent years. However, the accuracy rates of previous methods are not high. The reason is that the manifold of face image data is not utilized sufficiently and some patitcular characters of the individual image are neglected in these methods. Thus a new method to form graph of data is proposed in this paper and is used to develop two face recognition algorithms. The maximum minimum value of manifold can be preserved based on the new graph. At the same time the pixels correlation in individual image is considered sufficient under the constrain of spatial smoothness in the two developed algorithms. Therefore, the right recognition rates are enhanced by the two proposed algorithms. This is further confirmed by experiments.