Semantic based retrieval system of arctic animal images

In this paper we propose a semantic based image retrieval system in the domain of arctic animals. The proposed system exploits a semantic engine capable of adapting the processing steps both to the users' need and to the arctic image domain. This exibility has been achieved by three main steps: 1) arctic domain ontology modeling, 2) identification of features peculiar of the images we are dealing with and, 3)interface composition to support user interaction and search customization. The performance of the proposed system was tested using a set of 200 images depicting wild animals living in polar environment while users performed different search tasks specifying different constraints through the user interface. The results show both the accuracy of the retrieved results and the exibility to the users' constraints of the proposed system.