Modeling, Indexing and Retrieving Images Using Conceptual Graphs

When dealing with the complexity of an image as part of the indexing process, keywords are not sufficient to obtain an index that is a faithful representation of the image content. We propose to use the conceptual graphs formalism as the indexing language, which allows to use not only keywords, but also relations between them. The obtained indexes are more precise, and retrieval effectiveness is thus improved. Our paper presents a system that provides a computer-assisted image indexing process, which is performed according to a formal image model. The result of the indexing process, which is a set of conceptual graphs, is then organized so that to improve retrieval execution times. Our image retrieval system, called RELIEF, is implemented on an object-oriented DBMS and is available on the Web. It ensures the management of an image test collection and gives good results, with respect to both execution time and quality of answers.