Modified similarity metric for non-local means algorithm

The non-local means (NLM) algorithm exploits the self-similarity property of images to suppress noise. Similarities are measured by neighbourhood comparisons; however, the original NLM algorithm uses only translational neighbourhoods, which do not fully exploit the symmetric characteristics that exist in many images. Several neighbourhood transformations are introduced for similarity comparison so that the self-similarity property of images can be better exploited. Experimental results show that, compared with the original algorithm, substantial PSNR improvements can be achieved when appropriate neighbourhood transformations are used in the similarity comparison.