The main goal of IMEDIA is to develop content-based image indexing techniques and interactive search and retrieval methods for browsing large multimedia databases by content. We study ” generic ” image databases (e.g. the web) as well as ” domain-specific ” databases (e.g. facial imagery, medical imagery…). These two categories are related respectively to image retrieval and object recognition. In fact, image retrieval is a wider domain that contains object recognition but also integrates user interactions. More generally, the IMEDIA project-team achieves research, collaborations, and technology transfer on the complex issue of intelligent access to multimedia data streams.

Research themes

• Image indexing. Choosing a suitable feature space and building efficient indexes (tradeoff storage cost/query cost).
• Interactive search in large image databases. Perceptual similarity, image matching, uncertainty modeling, user modeling, relevance feedback, partial queries (on regions of interet).
• Browsing by content. Information theory, clustering methods, probabilistic modeling, visual content overview, mental image search, user interface.
• Multimedia indexing. Generalization to heterogeneous data, text/image indexing, application to multimedia documents.

International and industrial relations

• Industrial collaboration with Thales, Sagem, Philips, France Telecom, TF1, Thomson Multimédia, Police Judiciaire, INRA, l’IRD etc.
• International collaboration within european networks of excellence Delos 2 and Muscle, and the integrated project Acemedia, EPFL (Switzerland), NII(Japan), SupCom (Tunisia) etc.
• Participation to the RNRT RECIS project (Searching Images and Sound by Content), projet PRIAMM MediaWorks, ACI “Masses de données” Biotim