Robust Sparse 2DPCA and Its Application to Face Recognition

This paper proposes robust sparse 2DPCA (RS2DPCA) that makes the best of semantic, structural information and suppresses outliers. The proposed RS2DPCA combines the advantages of sparsity, 2D data format and L1-norm. To assess the performance of RS2DPCA in face recognition, experiments are performed on two famous face databases, i.e. Yale, and FERET, and the experimental results indicate that the proposed RS2DPCA outperform the same class of algorithms, such as RSPCA, 2DPCAL1.