Query-sensitive similarity measure for content-based image retrieval using meta-heuristic algorithm

Content based image retrieval (CBIR) systems retrieve images linked to the query image (QI) from enormous databases. The feature sets extracted by the present CBIR systems are limited. This limits the systems' effectiveness. This study extracts expansively robust and important features from the images database. These features are then kept inside the feature repository. This feature set is comprised of color signature containing features of shape and color. Here, from the given QI, features are extracted in the same manner. Accordingly, new evaluation of similarity employing a meta-heuristic algorithm (genetic algorithm with Iterated local search) is conducted between the query image features and the database images features. This study proposes CBIR system that is evaluated by investigating the number of images (from the test dataset). Meanwhile, the system's efficiency of is assessed by performing computation on the value of precision-recall for the results. The obtained results were better in comparison other advanced CBIR systems in terms of precision.