Fuzzy Multilevel Image Thresholding Based on Improved Coyote Optimization Algorithm

Abstract

Due to the computational complexity of multilevel image thresholding, Swarm Intelligence Optimization Algorithm (SIOA) has been widely applied to improve the calculation efficiency. Therefore, more and more attention has been paid to exploring the application of the latest SIOA in multilevel segmentation. This article takes Otsu and fuzzy entropy as the objective functions, using Coyote Optimization Algorithm (COA) for multilevel thresholds optimization selection, through fuzzy median aggregation of local neighborhood information and then forms the Fuzzy Coyote Optimization Algorithm (FCOA), so that the thresholding image segmentation can be achieved in the end. To prevent the COA algorithm from falling into the local optimum, this article follows the differential evolution strategy adopted by the standard COA, using the number of iterations to construct the differential scaling factor to form the Improved Coyote Optimization Algorithm (ICOA). The experimental results show that fuzzy Kapur entropy and fuzzy median value aggregation-based ICOA(FICOA) achieves better image segmentation quality. Compared with Grey Wolf Optimizer (GWO), Fuzzy Modified Quick Artificial Bee Colony and Aggregation Algorithm (FMQABCA) and Fuzzy Modified Discrete Grey Wolf Optimizer and Aggregation Algorithm (FMDGWOA), FCOA and FICOA have certain advantages in visual effects of image segmentation and PSNR, FSIM evaluation indices. Particularly compared with GWO (also a wolf evolutionary algorithm), FICOA shows significant advantages.

Keywords

• Coyote optimization algorithm
• information entropy
• image segmentation
• multilevel thresholding