An image segmentation method based on improved Monarch Butterfly Optimization

Image segmentation refers to splitting an image into non-identical and significant parts for more accurate classification or interpretation. In general, multi-level thresholding methods are used to find the best threshold levels for image segmentation. Although researchers have proposed different techniques to find the optimal thresholds, this issue is still considered open in the image processing. This paper presents an improvement for the Monarch Butterfly Optimization (MBO) algorithm to find the optimal threshold values using between-classes Otsu variance, and we call it as improved MBO “IMBO”. We define a new adaptive crossover rate and change the method of updating the butterflies for enhancing migration and adjusting operators of the MBO algorithm. The efficiency of the proposed method is analyzed for the determination of the optimal threshold on eight benchmark images, and the results are compared with the values obtained from genetic algorithm (GA), particle swarm optimization (PSO), MBO, and modified bacterial foraging (MBF) algorithms in terms of PSNR and SSIM values. The results show the superiority of the IMBO algorithm. The efficiency of the proposed algorithm is evaluated on the optimization of 20 benchmark functions with dimensions 2 and 20. The proposed method showed the best performance compared to GA, ant colony optimization (ACO), population-based incremental learning (PBIL), PSO, and MBO algorithms on 16 benchmark functions.