Non local means algorithm with adaptive isotropic search window size for image denoising

The selection of search window size is a critical issue in Non-local means (NLM) algorithm. The selection of small search window size for a pixel lying in smooth or homogeneous region limits the denoising effect due to contribution of only a small number of related pixels, whereas the selection of large search window size for a pixel lying in non-smooth or transition region degrades the denoising quality due to contribution of more and more unrelated pixels. The proposed algorithm selects the optimal search window size for each pixel, based on the gray level difference which indicates the characteristics of the region. Experimental results on standard images show that the proposed algorithm outperforms the conventional NLM algorithm in terms of PSNR and visual quality. It also preserves the image details such as edges and texture in an image at higher noise level.