Semantic-based image retrieval: A fuzzy modeling approach

Abstract

In this paper, we propose a new fuzzy based image retrieval approach to reduce the semantic gap in content-based image retrieval systems. Our main contributions are: (1) an algorithm for reduction of feature space dimensionality, (2) a fuzzy modeling approach to model the expert human behavior in the image retrieval task, (3) a fuzzy system for semantic-based image retrieval, and (4) a training algorithm for creating the fuzzy rules. The proposed solution not only is a novel idea in the semantic-based image retrieval field, but also has enough potential in learning semantics from users and making a powerful approach to improve the performance of CBIR systems, as the results of our experiments on a set of 2000 images support our claim.