General and invariant image retrieval from visual databases based on multiple search strategies

This paper is related to visual information systems in general, and to information retrieval from image databases in particular. We propose an approach based on the fusion of retrieval results returned by multiple retrieval strategies to improve the effectiveness of image retrieval from image databases. Multiple retrieval strategies proposed and used in this paper are multiple content representations and multiple queries. We consider the case of textures and consider both the general retrieval case as well as the case of invariant image retrieval. Experimental results show, through benchmarking on two image databases, that retrieval effectiveness is improved in a very appreciable way with the fused model.