Effective management and exploitation of multimedia documents requires extraction of the underlying semantics. Multimedia analysis algorithms can produce fairly rich but imprecise information about a multimedia document. In this paper, a multimedia reasoning architecture is presented using the fuzzy extension of expressive SHIN, f-SHIN. First a segmentation algorithm generates a set of oversegmented regions and a classification process is employed to assign those regions with semantic labels. A semantic based refinement of the segmentation is follows and this information initializes the ABox of a fuzzy-knowledge that is used for multimedia reasoning. The proposed approach was tested on outdoor domain and shows promising results.
2nd International Workshop on Semantic Media Adaptation and Personalization, IEEE, London, UK, pp.44-49, December 2007.
[ Bibtex ] [ PDF ]