An efficient multilevel thresholding image segmentation through improved elephant herding optimization

This paper proposed an improved version of recently developed swarm-based metaheuristic algorithm elephant herding optimization (EHO) called Improved Elephant Herding Optimization (IEHO). In this IEHO, the opposition-based learning (OBL) rule and chaos-embedded sequences are incorporated with each iterative stage of EHO to maintain the proper balance between the exploration and exploitation phase. It regulates the movement of the search agents and avoids premature convergence. The effectiveness of the proposed model is evaluated in terms of finding the optimal threshold value in Multilevel thresholding (MTH) of image segmentation which separates the different objects of the images. The methods such as the Kapur entropy, Otsu and masi entropy are used as the objective function in this problem to determine the optimal threshold. The proposed IEHO’s performance is compared with the different variants of EHO, artificial bee colony (ABC) and artificial hummingbird algorithms (AHA). The simulation results regarding convergence speed, stability, and solution quality performance indicators, such as the structural similarity index (SSIM), feature similarity index (FSIM), and peak signal-to-noise ratio (PSNR) verify the viability of the above hybrid algorithm.