An Interactive and dynamic fusion-based image retrieval approach by CINDI

Abstract

This paper presents the image retrieval technique and the analysis of different runs of ImageCLEF 2007 submitted by the CINDI group. An interactive fusion-based search technique is investigated in both context and content-based feature spaces. For a context-based image search, keywords from associated annotation files are extracted and indexed based on the vector space model of information retrieval. For a content-based image search, various global and region-specific local image features are extracted to represent images at different levels of abstraction. Based on a user's relevance feedback information, multiple textual and visual query refinements are performed and weights are adjusted dynamically in a similarity fusion scheme. Finally, top ranked images are obtained by performing both sequential and simultaneous search processes in the multi-modal (context and content) feature space. © 2008 Springer-Verlag Berlin Heidelberg.