K-Means Segmentation of Underwater Image Based on Improved Manta Ray Algorithm.

Area of Science:

• Computer Vision
• Artificial Intelligence
• Optimization Algorithms

Background:

• Traditional K-means image segmentation suffers from randomness and local optima, reducing segmentation quality.
• Existing methods struggle to balance global and local optimization effectively.

Purpose of the Study:

• To propose an improved manta ray foraging optimization (IMRFO) method for K-means image segmentation.
• To enhance the stability and accuracy of K-means by addressing its inherent limitations.

Main Methods:

• Incorporating Lévy flight and random walk learning into manta ray foraging optimization.
• Integrating particle swarm optimization concepts to improve convergence accuracy.
• Evaluating IMRFO against 11 other algorithms on standard test functions and underwater images.

Main Results:

• IMRFO demonstrated superior optimization ability compared to 7 basic and 4 variant algorithms on test functions.
• Experiments on underwater images showed improved Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index Measure (SSIM), and Feature Similarity Index Measure (FSIM) using IMRFO.
• The optimized K-means method exhibited enhanced stability and performance.

Conclusions:

• The proposed IMRFO significantly improves K-means image segmentation by enhancing global and local optimization capabilities.
• IMRFO offers a more robust and accurate approach to image segmentation, particularly for complex datasets like underwater imagery.