LEARNING INTERACTION MEASURE WITH RELEVANCE FEEDBACK IN IMAGE RETRIEVAL

Relevance feedback is an eective approach to bridge the gap between low-level featureextraction and high-level semantic concept in content-based image retrieval (CBIR). In this paper,we further improve the use of users feedback with multi-feature query and the Choquet integral.Taking into account the interaction among feature sets, feedback information are used to adjust thefeature's relevance weights that are considered as the fuzzy density values in the Choquet integralto dene the overall similarity measure between two images. The feature weight adjustment andintegration aims at minimizing the dierence between users desire and outcome of the retrieval system.Experimental results on several benchmark datasets have shown the eectiveness of the proposedmethod in improving the quality of CBIR systems.