ROBUST ESTIMATOR BASED MEDIAN FILTER FOR IMPULSE NOISE SUPPRESSION FROM DIGITAL IMAGES

Median filter and its several variants form the major representatives of the family of nonlinear filtering techniques and have been proved quite effective in the elimination of ‘non Gaussian noise’ such as ‘impulsive’ along with the ability of well preserving the edge details of an underlying image at lower noise density. The main limitation with the nonlinear median filter is that the healthy pixels are altered and the fine details of the image are smoothed out during the restoration stage of a highly corrupted image. To meet the requirement of obtaining a clean image with the important information of the noisy images preserved intact, a novel, robust estimator based, decision type nonlinear median based algorithm for the suppression of impulsive noise from the heavily noised digital images has been proposed in this paper. The proposed Robust Estimator based Median Filter (REMF) is implemented on a MATLAB platform and the simulation results obtained for the images contaminated with fixed value impulsive (or salt and pepper noise (SPN)) noise show that the proposed scheme performs effective identification and removal of the high density impulse noise while retaining the originality of the image as compared to the standard median filter (SMF), Adaptive Median Filter (AMF), Novel Median based Filter (NMF) and many other filters. Proposed filter is also tested on a large number of images contaminated with random valued impulsive noise (RVIN) of varying noise densities, on the images corrupted with additive white Gaussian noise (AWGN) and also on the images affected by mixed Gaussian-impulsive noise (MGIN). Simulation results prove that the proposed algorithm effectively suppresses the high density salt and pepper noise (SPN), medium density random valued impulsive noise (RVIN), low density Gaussian noise (AWGN) and mixed Gaussian and impulse noise (MGIN) of extremely lowest proportion.