An improved nonlocal means algorithm based on regression analysis

In this paper we propose a denoising method under the powerful framework-non local means (NL-means), in conjunction with regression analysis. First, the conception and development of NL-means is proposed. Second, an image important area map (IIAM) is calculated for the protection of edge and structure regions during denoising. Therefore, improved version of NL-means is carried out, which uses a novel patch similarity calculation rule benefited from IIAM. This algorithm can better extract and smooth the important information in the noisy image and finally, experimental results validate the proposed algorithm.