Abstract

Image segmentation remains a fundamental challenge in computer vision, where standard optimization algorithms such as Particle Swarm Optimization (PSO) often suffer from premature convergence and loss of population diversity, which limits segmentation accuracy. This study integrates chaotic maps into the PSO algorithm to improve population diversity, prevent premature convergence, and enhance segmentation quality. The aim is to evaluate three chaos-enhanced PSO (CE-PSO) variants, Logistic+PSO, Sinusoidal+PSO and Circle+PSO, for image segmentation using the Apple Image Dataset obtained from Kaggle. Chaotic sequences were used in place of the pseudo-random number generators employed for swarm initialization, parameter adaptation, and particle-position updates. The models were implemented in Python and evaluated on the preprocessed Apple Image Dataset using PSNR, SSIM, FSIM, convergence speed, and computational time, averaged over 30 independent runs per algorithm. Sinusoidal+PSO achieved the best results, with a PSNR of 27.89 ± 0.98 dB, SSIM of 0.9123 ± 0.018, and FSIM of 0.8967 ± 0.021, followed by Circle+PSO (27.12 ± 1.05 dB, 0.9034 ± 0.019, 0.8856 ± 0.023) and Logistic+PSO (26.78 ± 1.12 dB, 0.8912 ± 0.021, 0.8745 ± 0.025). Sinusoidal+PSO also converged fastest, reaching its solution in an average of 41 iterations, while Circle+PSO required the least computational time (2.38 ± 0.21 s). The study establishes that Sinusoidal+PSO offers the best accuracy for high-precision applications such as agricultural image analysis and automated fruit inspection, Circle+PSO offers the best computational efficiency for real-time, time-critical vision systems, and Logistic+PSO offers a computationally simple, general-purpose baseline.