An Image Thresholding Method Based on Differential Evolution Algorithm and Genetic Algorithm

Abstract

Thresholding is the simplest but most effecttive segmentation technique for image analysis. However, the computational complexity increases exponentially with the increase of threshold number in order to seek the most appropriate threshold values. Therefore, stochastic optimization algorithm are often used to overcome excessive computational problems, but the single optimization algorithm often falls into the local optimum. In general, hybrid algorithm is able to produce better performance. As a result, a parallel coupled mode(DE_GA in brief) of differential evolution algorithm (DE) and genetic algorithm (GA) is proposed for solving multi-threshold problem and The maximum variance is used as the fitness function. The experimental result displays that compared with a single algorithm, the results of the hybrid algorithm are relatively stable, which means that the parallel coupled DE_GA algorithm combined with Otsu might be an effect and practical image segmentation method.