Feedback-based image retrieval using probabilistic hypergraph ranking augmented by ant colony algorithm

Abstract

One fundamental issue in image retrieval is its lack of ability to take advantage of relationships among images and relevance feedback information. In this paper, we propose a novel feedback-based image retrieval technique using probabilistic hypergraph ranking augmented by ant colony algorithm, which aims at enhancing affinity between the related images by incorporating both semantic pheromone and low-level feature similarities. It can effectively integrate the high-order information of hypergraph and the feedback mechanism of ant colony algorithm. Extensive performance evaluations on two public datasets show that our new method significantly outperforms the traditional probabilistic hypergraph ranking on image retrieval tasks. © Springer-Verlag Berlin Heidelberg 2013.