Advancements in Content-Based Image Retrieval Using Visual Features and Color Descriptors

Abstract

Content-Based Image Retrieval (CBIR) has become an essential approach for retrieving visually similar images from large-scale databases by utilizing intrinsic image features rather than relying on textual annotations. This study presents a comprehensive analysis of CBIR systems, focusing on their architecture, feature extraction strategies, and similarity-based decision mechanisms. The framework consists of two fundamental modules: feature extraction and decision-making, which together enable efficient retrieval based on visual characteristics. The study further examines different query paradigms, including keyword-based, visual, query-by-example, and trace-based approaches, highlighting their limitations and usability challenges. In addition, the evolution of image retrieval systems from keyword-based methods to content-based and semantic-based techniques is discussed. Special emphasis is given to image characterization using color descriptors, including RGB and HSV color spaces, color quantization, and computational descriptors such as histograms, color moments, and dominant color representation. The results indicate that while low-level features such as color, texture, and shape are effective for image representation, they are insufficient to fully capture high-level semantic meaning. The semantic gap between machine interpretation and human perception remains a major challenge. Future research should focus on integrating deep learning and semantic modeling techniques to enhance retrieval accuracy and system performance