Improved Northern Goshawk Optimization Algorithm for Medical Image Segmentation

The computational complexity of multi-level medical image segmentation will continue to increase with the increase of the number of thresholds. Therefore, in this paper, we propose improved Northern Goshawk Optimization based on Levy and Simulated Annealing (SA) to Northern Goshawk Optimization (LSANGO) and optimize the fuzzy Kapur objective function for medical image segmentation. In the improvement of the LSANGO, the Levy flight strategy was firstly added in the exploration stage, and its global search ability was enhanced by using the random jump characteristics of the step size of this strategy, thereby obtaining a faster convergence speed. Then, in the update position stage, the SA mechanism is introduced, which effectively solves the problem that the Northern Goshawk Optimization (NGO) tends to fall into local optimum during the optimization process. To better evaluate the performance of the improved algorithm, the LSANGO is compared with NGO, Fuzzy Coyote Optimization Algorithm (FCOA), Whale Optimization Algorithm (WOA) and Altruistic Harris Hawks’ Optimization Algorithm (HHO_Altruism). Finally, the superiority of LSANGO is verified by quantitative analysis of peak signal-to-noise ratio (PSNR) and feature similarity (FSIM) performance indicators.