A good neighbor is a great blessing: Nearest neighbor filtering method to remove impulse noise

Abstract

Impulse noise is one of the common noise types that affect images. Median filtering denoising method has been widely used for impulse noise. However, as the noise density increases beyond 50%, the noise reduction quality of median filter reduces sharply. In this paper, a very simple nearest neighbor filtering method11 The MATLAB code for the proposed NNFM algorithm is given at: https://github.com/rafimiet/NNFM.git is proposed for impulse noise reduction. The proposed method exploits the inter-pixel correlation in images. More spatially closer the pixels within an image, more the correlation amongst them. The filter window is partitioned into multiple clusters based on their distance from the pixel that needs to be corrected. The closest non-noisy pixel is searched and is used to replace the corrupted pixel. The proposed method has been simulated for grayscale as well as color images. The simulation results show that the proposed method outperforms the state-of-the-art methods. Also, the computational complexity of the proposed methods is very small as compared to the state-of-the-art denoising methods. The only method that outperforms the proposed method is median filtering method.

Keywords

• Impulse noise
• Denoising
• Nearest neighborhood
• Noise reduction
• Salt and pepper noise
• image filter;