An adaptive non-local means image denoising model

Non-local means (NLM) is an effective denoising method that explores self-similarities between neighborhoods in the image for noise removal. The traditional NLM method computes pixel similarity using the globally fixed decay parameter and invariable matching window. However, a fixed decay parameter and constant window size for the whole image is difficult to ensure that the NLM method can denoise effectively both edge pixels and smooth area. To address this problem, an improved method is proposed, which classifies the image into several region types, according to the region character, an adaptive decay parameter and local window is adaptively adjusted to match the local property of a region. The results of experiments demonstrate the adaptive NLM model denoise the image and retain the details more effectively than traditional NLM diffusion.