Content-Based Image Retrieval: The State of the Art

Content-Based Image Retrieval (CBIR) is the solution to the image retrieval problem based on the contents of the query image. The objective of the CBIR system is to retrieve the visually similar images from the database efficiently and effectively but still, no satisfactory performance has been achieved. The performance of the CBIR system mainly depends on the feature extraction, feature selection, distance measures (similarity computation), Classification, and ranking of matched images. Feature extraction is the procedure of deriving the set of features from images for matching the visual similarity and they can be further classified based on color, texture, and shape descriptors. Performance is not up to mark when Color, Texture or Shape descriptors individually applied. Better determination of blend of Color, Texture, and/or Shape features can enhance performance in the context of precision and recall. This paper mainly concentrates on the brief review of the different state of art techniques used for CBIR along with prerequisite knowledge over this domain.