Face Recognition Based The Improvement Glram Algorithm

Face recognition is an advanced technology which analyzes the faces to extract effective features by computer.This technology not only involves engineering disciplines such as pattern recognition, image process, but also involves non-engineering disciplines such as physiology, psychology. Because this technology needs to study some other factors as the face expression sometimes. Nowadays face recognition is one of the most popular research areas in the world.The algorithm, which is studied in this paper, is a subspace method. This algorithm is called generalized low rank approximations of matrices (GLRAM).GLRAM has many advantages compared with the traditional PCA algorithm. First GLRAM is a method by two-side reduction dimension. It can achieve a higher compression ratio when the reconstruction error is small. However, the efficiency of GLRAM is low because of iterative process. In this paper, we will use the solution of the algorithm as key point and reference solving process of 2DPCA algorithm, proposed limit GLRAM algorithm. Limit GLRAM algorithm improves the efficiency of traditional GLRAM algorithm. Next this paper proposes a method of face recognition combining limit GLRAM with PCA by analysis of the property of limit GLRAM. This algorithm gets further optimized and this paper fully proved that the limit GLRAM algorithm is superior to traditional GLRAM algorithm by large number of experiments. At last, we find the dimension of GLRAM subspace will not affect the final reconstruction result when we use GLRAM+PCA algorithm. This algorithm's application ability get further improvement based on this property of GLRAM+PCA algorithm.