Sparse representation based classification performance under different optimization forms for face recognition

Sparse representation-based classification (SRC) has become one of the most powerful methods for robust face recognition. However, there are some limitations of SRC performance at the presence of noise, occlusion, and illumination variation problems, which make it unstable. Therefore, we investigate the performance of SRC under different data conditions by applying the most powerful optimization methods based on SRC and focusing on the corrections between data samples and the sparseness. For evaluation, we utilize several challenging face datasets that include diversity of illumination and occlusion conditions.