Improved non-local mean denoising algorithm based reliability of reference pixel

Abstract

In this paper, we propose a method for improved the non-local mean denoising algorithm by using the similarity in comparison with reference pixels. The noise in image is eliminated by the calculation to follow values as the patterns are similar in non-local mean denoising alogirthm. In the conventional algorithm, all the comparison results of the pixels are included in the denoising process, and even the low similarity result is included. The method causes the denoised effect to incorrect. In this paper, we propose an algorithm that reduces the influence of low-similarity reference pixels or improves the effect of high-similarity reference pixels. The proposed method improves both the object quality and the subject quality.