SNLM: A switching non-local means filter for removal of high density salt and pepper noise

Abstract

A Switching Non-Local Means (SNLM) filter is presented for high-density salt and pepper noise reduction. Firstly, the impulse noises are detected, based on the fact that their values must be the extreme gray-level of the image. Then, at the filtering stage, the noise-free pixels remain unchanged and noisy pixels are restored using a modified non-local means filter. However, to calculate the weights of the filter, only noise-free pixels are considered. It means that in a search window around the noisy pixel, some small patches are taken into account around noise-free pixels and the similarity between these patches and the central patch determines the weights. Experimental results show that the proposed method can provide better performance than many of the existing impulse denoising methods in high-density impulse noise in terms of PSNR, and MAE.