Unified multimedia annotation system using MPEG-7 visual descriptors and semantic event-templates

Multimedia contents with spam tags degrade the quality of the result of multimedia search. Therefore, the metadata related to the content are needed to improve the search results. This paper proposes an efficient multimedia annotation system that not only describes visual information by MPEG-7 visual descriptors but also semi-automatically generates the semantic information by using LSCOM ontology. Especially, we make use of the event-templates which predefine the relations among the events, characters and objects in scenes so that the users can easily create the scene metadata.