Image Clustering with Optimization Algorithms and Color Space.

Area of Science:

• Computer Vision
• Image Processing
• Machine Learning

Background:

• Image clustering requires high cluster homogeneity, typically achieved in grayscale segmentation via multiple thresholds.
• Conventional thresholding algorithms are inadequate for color image segmentation.
• Determining optimal multiple thresholds is a significant challenge.

Purpose of the Study:

• To introduce a new color image clustering algorithm utilizing multilevel thresholding.
• To demonstrate the application of multilevel thresholding techniques in color image clustering.
• To improve the homogeneity and efficiency of color image segmentation.

Main Methods:

• Threshold selection techniques (Otsu, Kapur) applied to individual color channels.
• Integration of objective functions with Forest Optimization Algorithm (FOA) and Particle Swarm Optimization (PSO).
• Color space division into smaller prisms/cubes using determined thresholds, with prism volume impacting cluster homogeneity.

Main Results:

• The proposed algorithm effectively clusters color images using multilevel thresholding.
• Optimization algorithms (FOA, PSO) enhance threshold determination for better segmentation.
• Reduced sub-cube volumes through multiple thresholds lead to improved cluster homogeneity.

Conclusions:

• The presented algorithm offers an efficient approach to color image clustering.
• Multilevel thresholding, guided by optimization algorithms, is effective for color image segmentation.
• The method demonstrates superior performance compared to conventional image clustering techniques.