A Novel Correlated Microstructure Elements Descriptor for Image Retrieval

In recent years, substantial progress has been made in developing new descriptors to enhance content-based image retrieval (CBIR) systems. These advancements often focus on leveraging the relationship between low-level features such as color and texture. This study introduces the Correlated Microstructures Elements Descriptor (CMED), a novel descriptor that integrates three low-level features to improve image retrieval performance. Our experiments on three distinct natural image datasets reveal that CMED significantly outperforms both classical and state-of-the-art descriptors. The proposed algorithm demonstrates superior indexing and retrieval capabilities, achieving up to 26.41% improvement compared to the MPEG-7 standard and 10.75% compared to contemporary state-of-the-art descriptors. The findings underscore CMED's potential to advance the field of CBIR, offering robust solutions for accurately retrieving images based on semantic content.