Annotation of multimedia data using semantic metadata

A user of a digital camera produces lots of images throughout a year and saves them to his local hard disk. After some time the amount of photos exceeds the critical mass manageable without specialized tools. Most people create an intuitive structure for storing their personal image library. They create folders for images that are taken in the same context, for example "Photos from the IFOST 2006 Forum", however this does not enable the user to find a photo which shows a specific person, object or even expresses a specific idea or feeling when needed. In this paper a new approach to metadata production is presented. For this purpose, a new interactive tool for multimedia content acquisition and classification has been developed. The user can decompose a given multimedia content into units and easily annotate each until adding basic information such as place, time, locate, etc. as well classification information such as even type, relationship type, etc. according to the MPEG-7 standard. At the end of this production, the tool automatically produces a semantic description of the overall set of the annotated units. The new idea proposed in this work is to combine the intrinsic semantics of each annotated unit with implicit semantic information derived from the structural description, hence reducing the needs to perform complex signal processing operations on the content.