GA based Denoising of Impulses (GADI)

Abstract

In this paper we have proposed a novel method of removing random valued impulse noises from the digital images. A variable window such as 5 × 5 and 3 × 3 are utilized for such purpose. The proposed method is a switching median filter. The detection of noises in every 5 × 5 window of the noisy image is done using all neighbor directional weighted pixels. After detection of noisy pixel in the 5 × 5 window the proposed filtering scheme restored it to a pixel which is most suitable in the 3 × 3 and 5 × 5 window regions. This scheme is based on weighted median filtering on the 3 × 3 window regional pixels. Three user parameters of the proposed noise removal operator are searched in a 3D space using a randomized search and optimization technique i.e., Genetic Algorithm. Implementation of the scheme shows better noise removal performance and also preserves the image fine details well. © 2011 Springer-Verlag.