Significance of dictionary for sparse coding based face recognition

Abstract

Sparse representation based classification (SRC) successfully addresses the problem of face recognition under various illumination and occlusion conditions, if sufficient training images are given. This paper discusses the significance of dictionary in sparse coding based face recognition. We primarily address the problem of sufficiency of training data in various illumination conditions. The dictionary is generated using a lower dimensional representation of image, which emphasizes the subject specific unique information of the face image. This representation is called weighted decomposition (WD) face image, because it attempts to give more weightage to unique information of face image. The effect of illumination in computation of WD face image is reduced using edginess based representation of image, which is derived using one-dimensional (1-D) processing of image. 1-D processing provides multiple partial evidences, which are combined to enhance the face recognition performance. The experimental results suggest that the proposed approach addresses the issue of sufficiency of training data efficiently.