Novel Chaotic Elephant Herding Optimization for Multilevel Thresholding of Color Image

Abstract

Elephant herding optimization (EHO) is a newly developed metaheuristic algorithm which is inspired from the herding behavior of elephant in nature. However, slow convergence is the main disadvantage of the basic EHO algorithm. To improve the global convergence speed as well as the performance of the basic EHO, we propose a new variant of elephant herding optimization by introducing the chaos theory into it. This new variant of EHO algorithm is called chaotic elephant herding optimization algorithm (CEHO). The CEHO algorithm uses a set of chaotic maps that generate chaotic numbers for tuning the control parameters of the basic EHO. The chaotic maps generate different sets of non-repetitive random numbers in a specified range, which are suitable for increasing the searching domain of the algorithm. The performance of the proposed CEHO is applied to a set of images collected from “Berkeley segmentation dataset” to find the optimal threshold values for multilevel image thresholding. The performance of the proposed algorithm is compared with the basic EHO, cuckoo search (CS), and artificial bee colony (ABC) quantitatively and qualitatively. The simulation outputs show that the proposed algorithm supersedes the others.