An Enhanced Approach to Content-based Image Retrieval using Harris Hawks Optimization and Automated Image Captioning
Roshni Padate, Ashutosh Gupta, Prąsun Chakrabarti, Arvind Sharma · 2024
This paper introduces an innovative approach to Content-Based Image Retrieval (CBIR) that leverages Harris Hawks Optimization (HHO) to improve feature selection and retrieval accuracy. CBIR systems are increasingly important in fields such as medical imaging and remote sensing, where efficient and accurate image retrieval is essential. Inspired by the cooperative hunting strategies of Harris hawks, the HHO algorithm efficiently balances exploration and exploitation, optimizing the feature extraction process to enhance retrieval outcomes. The proposed system is evaluated on various datasets and compared to other optimization techniques, such as Particle Swarm Optimization (PSO) and Genetic Algorithms (GA). The findings show that the HHO-optimized system outperforms these methods in terms of precision, recall, and F1-scores, while also reducing retrieval time.Additionally, the incorporation of image captioning into the CBIR framework further improves the user experience. By integrating Computer Vision (CV) and Natural Language Processing (NLP), the system generates descriptive captions along- side retrieved images, offering valuable textual context. This capability allows users to quickly evaluate the relevance of retrieved images, particularly in complex fields like medical imaging. Experimental results demonstrate that the combination of HHO-based optimization and image captioning significantly enhances both retrieval accuracy and user satisfaction, making it a powerful solution for large-scale image retrieval tasks.