To address the shortcomings of the information acquisition optimizer (IAO)—specifically its susceptibility to premature convergence, insufficient exploitation capability during later stages, and population diversity decay when applied to complex optimization problems—this paper proposes a multi-strategy improved information acquisition optimizer (MIIAO). Centered on balancing exploration and exploitation capabilities during the search process, this method incorporates several key strategies: an adaptive differential perturbation factor is designed to dynamically adjust the search step size; an elite-guided information acquisition mechanism is introduced to enhance convergence efficiency within high-quality regions; a diversity-based restart perturbation strategy is integrated to mitigate the risk of entrapment in local optima; and a mirror boundary handling technique is adopted to bolster the resilience of solutions near boundaries and improve the effectiveness of searching within the feasible domain. To validate the efficacy of the proposed method, MIIAO was applied to the CEC2014, CEC2017, and CEC2022 benchmark test suites and systematically compared against various representative intelligent optimization algorithms. Furthermore, the method was applied to multi-threshold image segmentation tasks based on Otsu’s criterion. Experimental results demonstrate that MIIAO consistently exhibits superior solution accuracy, convergence speed, stability, and statistical ranking across various dimensions and a diverse range of complex test functions; the results of the Wilcoxon rank-sum test and Friedman mean ranking further substantiate its comprehensive performance advantages. In the image segmentation experiments, MIIAO achieved superior Otsu objective function values across multiple test images and under various threshold settings, while also demonstrating higher segmentation quality and greater robustness across evaluation metrics such as PSNR, SSIM, and FSIM. In summary, the proposed MIIAO effectively enhances the original IAO’s global search capability, local exploitation capability, and ability to maintain population diversity, thereby demonstrating significant potential for practical application in both numerical optimization and multi-threshold image segmentation tasks.
Zhang et al. (Thu,) studied this question.