부산대학교 도서관

Content-based image retrieval (CBIR) is a critical domain in computer vision, dedicated to extracting images from databases based on their visual content rather than relying on text-based queries. Feature extraction plays a pivotal role in CBIR, converting image attributes like color, texture, and shape into numerical representations, facilitating efficient image matching. It lacks a comprehensive exploration of the distinct impacts of fusion methods and often falls short in detailing feature extraction architectures. This paper introduces a novel approach to tackle the proposing a hybrid feature extraction method. The method evaluates its performance using the Corel 1K dataset, a collection of 1,000 images spanning diverse categories, serving as a benchmark for assessing the efficacy of content-based image retrieval techniques in real-world scenarios. The results achieved by the proposed hybrid feature extraction method surpass those of existing models, with impressive precision (0.956), recall (0.872), and F -measure (0.892). This method is compared to the KMFO model and performance evaluation in CBIR using the SVM model. Future research is focus on refining and optimizing advanced feature extraction techniques to enhance the efficiency and effectiveness of CBIR systems.