Mining discriminative components with low-rank and sparsity constraints for face recognition