Image retrieval: a first step for a human centered approach

Image indexing using content analysis is known as a difficult task, involving the vision research domain. Using these tools in the context of a retrieval system is generally frustrating for users, due to a lack of interfaces development and to the difficulty for users to understand the low-level features managed by the system. We propose in this paper a general point of view for introducing a link between such systems and potential users. This includes image features based on visual perception models, a relevance feedback model, and a graphical interface to express the information need through user-system interactions.