Multi-class relevance feedback for collaborative image retrieval

Abstract

In recent years, there is an emerging interest to analyse and exploit the log data recorded from different user interactions for minimising the semantic gap problem from multi-user collaborative environments. These systems are referred as “Collaborative Image Retrieval systems”. In this paper, we present an approach for collaborative image retrieval using multi-class relevance feedback. The relationship between users and concepts is derived using Lin Semantic similarity measure from WordNet. Subsequently, the Particle Swarm Optimisation classifier based relevance feedback is used to retrieve similar documents. The experimental results are presented on two well-known datasets namely Corel 700 and Flickr Image dataset. Similarly, the performance of the Particle Swarm Optimised retrieval engine is evaluated against the Genetic Algorithm optimised retrieval engine.