An adaptive isotropic search window based NLM algorithm for image denoising

The non-local means (NLM) algorithm uses the self-similarity or repeated patterns present in images for denoising. NLM algorithm has been extensively researched due to its effectiveness and simplicity. In conventional NLM algorithm, the size of the search window is kept fixed for each pixel. Ideally, the search window size must optimally vary from region to region based on the characteristics of the search region. In this paper, we propose an adaptive NLM algorithm based on classification of homogeneous and heterogeneous regions using local entropy. The proposed algorithm selects an optimal search window size for each pixel based on region characteristics. The experimental results have shown that the proposed algorithm performs consistently better than the conventional NLM in terms of PSNR and visual quality for denoising the images at various noise levels.