A Hybrid Genetic Algorithm and Gravitational Search Algorithm for Image Segmentation Using Multilevel Thresholding

This paper presents a novel optimal multilevel thresholding algorithm for histogram-based image segmentation. The proposed algorithm presents an improved variant of the gravitational search algorithm (GSA), a relatively recently introduced stochastic optimization strategy. To strengthen its ability to achieve generation jumping when getting stuck at local optima, this paper proposes a novel algorithm, GA-GSA (genetic algorithm-based gravitational search algorithm) for image segmentation. In this paper, the proposed method employs both GA and GSA and the maximum entropy criterion as the objective function for achieve multilevel thresholding. To demonstrate the ability of the proposed algorithm, the novel method is employed on two benchmark images, and the performances obtained outperform results obtained using two other stochastic optimization methods, i.e., PSO (Particle Swarm Optimization) and GSA. The experimental results illustrate that the proposed algorithm could significantly enhance performance compared to other popular contemporary methods.