Structural smoothness low-rank matrix recovery via outlier estimation for image denoising

Hengyou Wang, Wen Li, Lujin Hu, Changlun Zhang, Qiang He

Abstract: Natural images often have intrinsic low-rank structures and are susceptible to interference from outliers or perturbation noise, especially mixed noise. Low-rank matrix recovery via outlier estimation (ROUTE) has been proposed to determine the location of gross corruption by estimating the outliers; however, this approach ignores local structural smoothness. In this paper, we incorporate TV norm regularization into the ROUTE model of low-rank matrix recovery, which is called SSROUTE. This model can ensure structural smoothness in image denoising that is vulnerable to outlier noise and additive white Gaussian noise simultaneously. In addition, to solve the reformulated optimal problem, we develop an algorithm based on the alternating direction method of multipliers. Experimental results show that the proposed algorithm achieves a competitive denoising performance, especially for mixed noise.