In this paper we present our research work on the identification of high-level concepts within multimedia documents through the introduction and utilization of contextual relations. A conceptual ontology is introduced, as the means of exploiting the visual context of images, in terms of high-level concepts and region types they consist of. A meaningful combination of these features results in a computationally efficient handling of visual context and extraction of mid-level characteristics towards the ultimate goal of semantic multimedia analysis. Evaluation results are presented on a medium-size dataset, consisting of 1435 images, 25 region types and 6 high-level concepts derived from the beach domain of interest.
3rd International Workshop on Semantic Media Adaptation and Personalization, IEEE, Prague, Czech Republic, December 2008.
[ Bibtex ] [ PDF ]