Image thresholding based on evolutionary algorithms

Abstract

The objective of this paper is to propose an adaptive-evolutionary method for thresholding which is used as an artificial intelligent algorithm for image segmentation especially for object segmentation. This method employs resistant versus mixed histograms because of its suitable fitness function selection that consists of the histogram details. As things develop in the paper, three evolutionary methods known as genetic algorithm (GA), imperial competitive algorithm (ICA) and adaptive particle swarm optimization are used to minimize the error function. Finally, a powerful algorithm for image thresholding is found. The comparisons and experimental results show that this system is better than other methods particularly Otsu’s, GA and even new algorithms like ICA. Key words: Segmentation, adaptive particle swarm optimization (APSO), genetic algorithm (GA), imperialist competitive algorithm (ICA), threshold, fitness function.