Removal of random valued impulse noise from grayscale images using quadrant based spatially adaptive fuzzy filter

Abstract

An efficient image restoration technique based on spatially linked-directional adjoining pixels and fuzzy logic for addressing moderate and highly corrupted grayscale images with Random Valued Impulse Noise is presented. The proposed technique decomposes a larger sized impulsive region of an image into numerous overlapping small patches for low as well as high density impulse noise estimation with enhanced image restoration results. This subdivision highlights the details (edges, lines and textures) present in the neighborhood and the locations of impulsive pixels are relocated in multiple regions. Direction-based fuzzy rules give appropriate reasoning for edge and texture detection in an image. A switching technique based fuzzified degree identifies a certain pixel of an image as a noise-free, noisy or edge pixel in the filtering phase. Instead of using a static threshold, a directional non-parametric approach is introduced that determines and sets the threshold adaptively which empowers the proposed filter to handle different types of images automatically. Extensive simulations are conducted on a large set of monochrome images contaminated with varying degree of impulse noise densities. Objective analysis performed using popular quantitative measures and subjective evaluation of the results shows the efficacy of the proposed filter over most of the bench-marked denoising filters.