Classification of objects and background using parallel genetic algorithm based clustering

Abstract

In this paper, a novel strategy based on the notion of threshold is proposed to accomplish segmentation of objects and background in a scene. Optimal threshold for two class and three classes problems are determined from the histogram of featured pixel values as opposed to the original normalized histogram. Genetic algorithm (GA) and Parallel Genetic Algorithm (PGA) based clustering algorithms are proposed to determine the optimal thresholds for two as well as three class problems. The optimal threshold could segment the noisy images. Our results, for two class problems, could be comparable with that of Otsu’s approach. Our approach yielded satisfactory results even for histograms having overlapping class distributions.