An Improved ADM algorithm for RPCA optimization problem

This paper presents an improved alternating direction method (IADM) algorithm for robust principal component analysis (RPCA) optimization problem. Firstly distortion compensation technique is employed to convert 2-D real nature image to the sparse approximation matrix. Secondly an improved Singular Value Decomposition (block-SVD) is presented to converge to the better value than traditional alternating direction method (ADM). Finally, reconstructed image is built up by sparse and low-rank matrix. To illustrate the effectiveness of proposed approach, several experiments are conducted. Experimental results show that, compare with SVT, ALM, APG and ADM, the proposed approach has faster rate and better performance.