Fruit-Fly optimization based feature integration in image retrieval

Area of Science:

• Computer Science
• Information Retrieval
• Artificial Intelligence

Background:

• Content-based image retrieval (CBIR) systems retrieve similar images using extracted features.
• Current CBIR methods use static feature weights, limiting results and increasing retrieval time due to full database searches.

Purpose of the Study:

• To improve the efficiency and performance of CBIR systems.
• To address limitations of static feature weighting and long retrieval times in existing CBIR approaches.

Main Methods:

• A novel two-level searching process was introduced for CBIR.
• The initial level employs an image selection rule to narrow down relevant images.
• The second level utilizes dominant color and radial difference patterns, with optimal dynamic weights assigned via the fruit fly optimization algorithm to color and texture features.

Main Results:

• The proposed two-level search framework significantly enhances retrieval performance.
• Dynamic weighting of color and texture features improves the accuracy of similarity measures.
• The system demonstrates reduced retrieval times compared to traditional CBIR methods.

Conclusions:

• The fruit fly optimization algorithm effectively assigns dynamic weights, optimizing CBIR performance.
• The two-level search strategy combined with dynamic feature weighting offers a more efficient and accurate image retrieval solution.