An image retrieval interface based on dynamic knowledge

Focus of this work is the presentation of an interface for an image retrieval system that makes users aware of the overall organization of the image database. The main features of this system are the multi-images query approach and the use of a dynamic knowledge base in order to improve the system performances. The multi-images query is utilized because often a single image does not fully cover the semantic query idea of the user, so, this work presents a system that allows the user to submit multi query images. Another relevant feature of the proposed system is the use of a dynamic knowledge based approach. This knowledge base is composed of two sets of clusters: the first one is obtained with an automatic clustering algorithm and the second one is built dynamically by the user when using the system. The results presented in this paper demonstrate that this approach provides accurate retrieval results showing acceptable interaction speed that can be compared with existing methods.