Image denoising using weighted nuclear norm minimization with multiple strategies

Low rank methods have shown to provide excellent denoising performance, of which weighted nuclear norm minimization (WNNM) is particularly effective. It assigns different weights to different singular values. However, it has limitations in three aspects, namely, no consideration of the noise effect on the similarity measure, a fixed feedback proportion of method noise, and an inflexible number of iterations for each image. In this paper, a general denoising framework based on WNNM is proposed that offers three strategies to mitigate the above drawbacks. The first strategy is to perform a coarse prefiltering on noisy patches before patch matching. The second is to adaptively feed different percentages of method noise back according to the additive noise levels. And the last strategy is to apply a stopping criterion based on Pearson's correlation coefficient during iteration. Experimental results demonstrate the efficiency of the proposed approach.