Metaheuristic Algorithms Applied to Color Image Segmentation on HSV Space.

Area of Science:

• Computer Vision and Image Processing
• Artificial Intelligence
• Bio-inspired Computing

Background:

• Image segmentation and edge extraction are crucial for image analysis.
• Unsupervised methods reduce the need for labeled training data.
• Metaheuristic algorithms offer robust optimization for complex problems.

Purpose of the Study:

• To propose an unsupervised method for color image segmentation and edge extraction in HSV space.
• To leverage metaheuristic algorithms (Firefly and Artificial Bee Colony) for improved image analysis.
• To enhance image segmentation accuracy and edge detection capabilities.

Main Methods:

• Phase 1: Pixel-based segmentation using the Firefly Algorithm (FA) and Gaussian Mixture Model (GMM) on individual HSV color channels.
• FA automatically determines the number of clusters based on histogram maxima.
• Phase 2: Region-based segmentation and edge extraction using the Artificial Bee Colony (ABC) algorithm on the grayscale image derived from segmented HSV components.

Main Results:

• Successful unsupervised segmentation of color images in HSV space.
• Effective edge extraction using the bio-inspired Artificial Bee Colony algorithm.
• Demonstrated the capability of metaheuristic algorithms in image processing tasks.

Conclusions:

• The proposed two-phase unsupervised method effectively performs image segmentation and edge extraction.
• The integration of FA and ABC algorithms provides a robust approach for color image analysis.
• This research contributes to advancements in unsupervised image processing techniques.