Weighted Schatten P-Norm Minimization with Local and Nonlocal Constraints for Noisy Image Completion

Abstract

Weighted Schatten p-norm minimization (WSNM) has been used successfully for noisy-free image completion. However, WSNM can introduce extra artifacts if the observed entries of image contain noise. In this paper, we present a novel WSNM-based method for noisy image completion, which incorporates both local smoothness and nonlocal self-similarity in a unified framework. More concretely, the analysis operator is utilized to ensure local smoothness and the nonlocal statistical modeling (NLSM) is adopted to constrain nonlocal self-similarity while WSNM is effective for completing the missing entries. To make the proposed method tractable and robust, the alternating direction method of multipliers (ADM-M) is employed to solve the above inverse problem. Experimental results show the effectiveness of the proposed method for noisy image completion.