Towards efficient image segmentation: A fuzzy entropy-based approach using the snake optimizer algorithm

• The study highlights the effectiveness of fuzzy entropy thresholding methods for image segmentation. • A novel Snake Optimizer (SO) is introduced to optimize multilevel threshold levels using fuzzy entropy. • The SO mimics snake mating behaviours to enhance optimization efficiency. • Experiments on ten test images demonstrate the method's superiority over existing techniques. • The Snake Optimizer achieves better fitness values, reduced CPU time, and higher quality metrics. Image segmentation is a critical aspect of image processing, particularly for applications requiring precise object identification. This study introduces a novel multilevel thresholding technique for grayscale image segmentation based on the Snake Optimizer (SO) algorithm, which is inspired by the mating behaviour of snakes. The proposed method aims to maximize fuzzy entropy to determine optimal threshold values, thereby enhancing segmentation accuracy. Extensive experiments were conducted on ten benchmark images, utilizing 2 to 5 thresholds, and comparing the performance of the SO algorithm against three state-of-the-art algorithms: Salp Swarm Algorithm (SSA), Grey Wolf Optimization (GWO), and Moth Flame Optimization (MFO).The results demonstrate that the SO algorithm consistently outperforms its counterparts across various metrics, achieving the highest average fitness values in 35 out of 40 attempts, with notable improvements in PSNR, SSIM, and FSIM. Specifically, the SO achieved an average PSNR of 26.617 for the 'Barbara' image at 5 thresholds, surpassing SSA (26.568), GWO (26.561), and MFO (26.515). Additionally, the SO recorded an SSIM of 0.9575 for the same image, significantly higher than SSA (0.9339), GWO (0.9114), and MFO (0.8934). The computational efficiency of the SO is also highlighted, with lower average CPU times, achieving 3.1217 seconds for the 'Ostrich' image at 2 thresholds, compared to SSA (3.1298), GWO (3.8539), and MFO (3.9343).Statistical analysis using Wilcoxon rank-sum tests further confirms the significant performance advantages of the SO algorithm, establishing it as a robust and effective solution for multilevel thresholding in image segmentation tasks.