A Novel Semantic-Based Image Retrieval Method

Abstract

In this paper, we design a fuzzy system for image retrieval to reduce the semantic gap in the content-based image retrieval systems. Our main contribution is three-fold: (1) designing a fuzzy modeling approach to model the expert human behavior in the image retrieval task, (2) a fuzzy system for semantic-based image retrieval, and (3) 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 has enough potential in learning semantics from the user and making a powerful approach to improve the performance of CBIR systems, as our experiments on a set of 2000 images supports our claim.