Multi-strategy remora optimization algorithm for color multi-threshold image segmentation

Area of Science:

• Computer Vision
• Artificial Intelligence
• Image Processing

Background:

• Multi-threshold image segmentation is crucial but computationally complex due to large search spaces.
• Existing optimization algorithms often suffer from local optima and slow convergence.

Purpose of the Study:

• To introduce the Multi-Strategy Remora Optimization Algorithm (MSROA) for efficient and accurate color image segmentation.
• To enhance optimization performance by preventing local optima and accelerating convergence.

Main Methods:

• MSROA integrates Beta random restarts with a "prior" property to avoid local optima.
• A random walk with fast predation and elite learning strategies are employed to boost convergence speed and accuracy.
• Performance evaluated on CEC2017 and CEC2020 benchmark suites and applied to Otsu's and Kapur's methods for image segmentation.

Main Results:

• MSROA demonstrated statistically significant improvements over seven state-of-the-art algorithms via Wilcoxon rank-sum tests.
• The algorithm accurately identified optimal threshold combinations, producing higher quality segmented images.
• Consistently higher PSNR, FSIM, and SSIM values indicate superior preservation of image details.

Conclusions:

• MSROA offers a robust and efficient solution for multi-threshold color image segmentation.
• The algorithm effectively balances exploration and exploitation for improved optimization.
• MSROA outperforms existing methods in segmentation accuracy and detail preservation.