Image Thresholding-Based Switching Filter for Salt & Pepper Noise Removal

Abstract

Switching filter is a popular kind of image salt & pepper noise removal techniques in recent years. Unfortunately, noise detection schemes in existing switching filters are usually unsuitable for dealing with those images containing a certain number of noise-free pixels with gray levels 0s or 255s. To alleviate this issue, a new noise detection scheme is proposed, which can be regarded as a generalization of MDBUTMF's (modified decision based unsymmetric trimmed median filter) scheme. The proposed detection scheme innovatively utilizes image thresholding results obtained by gray level extremes 0 and 255 to further exclude fake noise pixels. And MDBUTMF's noise restoration scheme is adopted to restore our detected noise pixels. A series of simulations demonstrate that the proposed algorithm has higher noise detection accuracy and better restoration effect with higher PSNR values.