ABSTRACT

In order to enhance the ability to capture a user's perception of image content on retrieving desired images from a large database and to suitably influence the parameters of a support vector machine (SVM) based on regional representation, the SVM learning process is integrated into a Gaussian firefly algorithm as a relevance feedback approach in region-based image retrieval (RBIR). The problem of image retrieval is formulated as a problem of optimization and is worked out by using SVM and the Gaussian firefly algorithm (FA). The optimization of the SVM classifier using a FA in a RBIR has faster convergence speed than the SVM and FA and also since the firefly algorithm is used in relevance feedback, only the top fireflies are used for comparison with database images, which reduces its computational complexity. The firefly algorithm is one of the latest artificial intelligence algorithms developed. Machine learning and relevance feedback methods have been offered to learn and to refine query concepts.