Removal of salt and pepper noise with non-uniform weighted mean partition

Abstract

Non-uniform Partition (NUP) and Directional Weighted Mean Filter (DWMF) are both effective salt and pepper denoising algorithms than many other common denoising algorithms, but some defects are still existing in the situation of high noise density thus reducing their denoising performance. In this paper, a novel salt and pepper denoising algorithm, Non-uniform Weighted Mean Partition, based on NUP and DWMF is proposed. In the situation of high noise density, the square effect appears after NUP denoising. DWMF has an effective detail restoration ability in salt-and-pepper denoising but fail to restore some noises in higher noise density. Therefore, in the proposed algorithm, DWMF is applied in different size of partition areas to reduce the incidence of residual noise thus fixing the problem in NUP detail restoration. The Least Square Approximation is used to restore the larger image sub-area while DWMF is used to restore the sub-area with smaller size. For the smooth areas, the Least Square Approximation has a good restoration ability while the DWMF can effectively restore the details in the complex area. The experimental results show that the proposed algorithm is better than the original NUP and DWMF denoising algorithms.