Hybrid min-median-max filter for removal of impulse noise from color images

Abstract

In this paper, an improved median based hybrid Min-Median-Max filter (M3F) has been proposed, to restore original image that is corrupted by impulse noise. The identification of the contaminated pixels is performed by local extrema intensity, i.e. using Min-Max noise detector. If any pixel is found corrupted, it will be changed by the resultant value of M3F algorithm, and uncorrupted pixels remain unchanged. Different color images have been considered to test the proposed method and better results have been found in terms of quantitative measures and visual perception. The presented algorithm can effectively reconstruct noise-free image from image which is corrupted with 70% noise level and also maintains the edges. Even up to 90% noisy image can be identified using proposed method. Experimental observations indicates that the proposed method removes high density impulse noise efficiently at high noise level and also keeps the originality of pixel’s value.