Bean Optimization Algorithm (BOA) is a kind of swarm intelligence optimization algorithms that draws inspiration from biomimetic concepts found in natural plant populations to tackle complex problems. BOA shows robust capabilities in global optimization and rapid convergence rates, making it applicable across diverse domains, including recovery, reconstruction, and target search. A central focus of this study involves enhancing and refining core search model, with the goal of bolstering its fine search performance. This research specifically addresses challenges related to fine-tuning BOA and extends its application to multi-threshold image segmentation. The improved BOA algorithm is employed for maximum entropy multi-threshold image segmentation, followed by a comparison with other intelligent optimization methods. Experimental results show the superior optimization precision and segmentation efficacy of the BOA.