Face Recognition Based on Sparse Representation and Joint Sparsity Model with Matrix Completion

In this paper we verify the impacts of the Joint Sparsity Model with Matrix Completion (JSM-MC) for the composition of training set in the context of face recognition using the Sparse Representation-based Classifier (SRC). A pre-processing step (histogram equalization) is performed in the face images to reduce the effects of illumination change. A clustering of training images is done to reduce the training set and uses the l
₁
-norm of the sparse representation coefficients instead of the residuals for classification. The results are evaluated using a database with different illumination conditions and we also investigate the behavior of the system when the face image is partially occluded.