Fixed 3x3 Threshold based Non Linear Filter for the Removal of Long Tailed noise and Artifacts in images

Year: 2012

Issue: Feb-Apr   

Title: Fixed 3x3 Threshold based Non Linear Filter for the Removal of Long Tailed noise and Artifacts in images 

Author Name: Vasanth Kishore, V. Jawahar Senthil Kumar 

Synopsis: 

A Threshold based fixed 3x3 filter for the removal of long tailed noise (impulse) and artifacts in images is proposed. The processing pixels are arranged in increasing order using a novel snake like sorting for finding median. The algorithm detects the corrupted pixel, if the absolute difference between the processed pixel and the unsymmetrical trimmed median is greater than a fixed threshold else termed as non noisy. Under high noise densities the computed median might also noisy. Hence the absolute difference of the median and unsymmetrical trimmed median filter is greater than second threshold then the corrupted pixels are replaced by the trimmed median of the current processing window else median is replaced else if condition fails the processed pixels is left unaltered. The filtering window is kept fixed at 3x3 for increasing noise densities. The proposed algorithm shows good results quantitatively and qualitatively when compared to existing and recent filters. The proposed algorithm is tested against different grayscale images and it gives higher Peak Signal-to-Noise Ratio (PSNR), low Mean square error (MSE) and good Image Enhancement Factor (IEF) with edge preservation capabilities even at very high noise densities. The proposed Algorithm is non adaptive and can replace several individual algorithms for the removal of different artifacts.