Discriminative Decomposition Structure-Sparse Representation for Face Recognition with Occlusion
HU Zheng-pin · Journal of Signal Processing · 2014
To solve the image recognition problem when existing occlusion,an algorithm combined Discriminative Decomposition( DD) model with structured sparse representation is proposed. First,images are decomposed to three parts, common component,low-rank condition component and sparse error component. Secondly,projection matrix on common component and low-rank component are computed respectively and the final projection matrix is obtained by fusing the two matrixes; Finally,the recognition step was constructed on the projection subspace using structured sparse representation. Experiment results on AR dataset prove our method perform better in recognition rate than BS( Block Sparse Representation),NS( Nearest Subspace) and SRC in low-dimension.