Abstract: To solve the image recognition problem with occlusion, we combined Discriminative Decomposition (DD) model with structured sparse representation. First, images are decomposed to three parts, common component，low-rank condition component and sparse error component; secondly, compute projection matrix on common component and low-rank component respectively and construct the final projection matrix by unite the two matrixes; Finally, the recognition was processed on the projection subspace using structured sparse representation. For our knowledge, it’s the first time to combine this composition method with structured sparse representation in PCA projection subspace. Experiment results on AR dataset verify our method can get higher recognition rate than BS (Block Sparse Representation),NS (Nearest Subspace) and SRC in low-dimension.