Face Recognition: A Sparse Representation-based Classification Using Independent Component Analysis

In this paper, we will describe a new method based on Sparse Representation-based Classification (SRC) for face recognition. We have used histogram equalization as a preprocessing method in order to overcome the illumination variation problem. Using Independent Component Analysis we have obtained a feature vector for each face image which is robust to illumination variations and occlusion. Although SRC is robust against occlusion, it is rather slow. By using features with smaller dimensions but enough information, we can obtain better recognition rates in shorter periods. This method was tested on Extended Yale B database and obtained the recognition rates of 98.51% and 95.77% in presence of 10% and 20% occlusion, respectively.