An Automated Image Segmentation Framework Using Fractional Calculus and Improved Pigeon Swarm Optimization with 2D Otsu Thresholding

Abstract

Research has proposed an automated image segmentation algorithm AIS-IPTSA-FC, which combines the U-Net active contour model enhanced by fractional calculus (FC), the improved traditional pigeon swarm algorithm (IPTSA), and the two-dimensional Otsu algorithm (2D-OA). This method has been validated on the SiftFlow and OASIS-3 datasets. The experimental results demonstrated that this hybrid algorithm achieved a segmentation accuracy of over 0.99 on the SiftFlow dataset, surpassing mainstream algorithms. In practical applications, this method performed well: in natural landscape images, the average values of structural similarity index (SSIM), peak signal-to-noise ratio (PSNR), and mean squared error (MSE) were 0.36, 10.79, and 5501.61, respectively. In medical images, the corresponding indicators were 0.0086, 6.84, and 7976.47, respectively. The above results demonstrate that the research method can effectively achieve image segmentation under complex conditions and provide a reliable foundation for the development of multi domain intelligent systems.