Four stage median-average filter for healing high density salt and pepper noise corrupted images

This paper introduces a novel four stage filter algorithm to ameliorate images corrupted by very high density salt-and-pepper noise. The proposed algorithm exhibits two parallel trimmed median filters (TMF) at the initial stage followed by a masking logic that selects denoised pixel based on the priority. To reduce the blurring effect, higher priority is given to TMF with small window size. In the absence of noise-free pixels, the current extreme pixel is left unchanged at the first stage. Further, the denoising of unprocessed extreme pixels is done with TMF of large size window at the second stage. The remaining noisy pixels are improved by the running average filter at the third stage. Finally, the last stage handles the noisy pixels at the boundary and rare scenario. Since the proposed filter utilized non-extreme pixels to estimate denoinsed pixels value, it effectively eliminates salt and pepper noise along with better edge preservation. The performance analysis of the proposed filter is carried out with various benchmark images for varying noise density. The experimental results demonstrate on an average improvement of 2.09 dB (0.018) and 1.06 dB (0.0478) of PSNR (SSIM) respectively for wide (10% - 90%) and very-high (90% - 98%) noise density ranges over state-of-the-art filters.