Facing a large amount of rich visual video information, conventional video search techniques such as fast forward/rewind are no longer sufficient. Users want to be able to browse, to be selective at what they see just like how they have accessed textual information. This creates a problem because raw video bits do not possess the same user-level information as text and thus are not directly search-able in the same way. Consequently, video needs to be retrieved and indexed through its semantic content represented in a swell structure manner. Thus the challenge is to provide ways of creating this well structured information effectively.
Today's technology in multimedia computing however, is a long way from providing solutions for fully automatic video content extraction. Thus we are exploring solutions of video annotation, providing tools to facilitate a human (the annotator), whose role is to annotate and extract video semantic content. The added semantic information allow more effective video information retrieval and management.
In this paper, we propose a hierarchical metadata model to represent video information. This model consists of two separate hierarchies of metadata. The first hierarchy is a directed acyclic graph, captures the relationship between video segments at the semantic level. The second hierarchy is a object composition graph, holds objects that represent meaningful content appearing in the video. We have developed a Video Metadata Authoring and Browsing System that uses video segmentation result and generates the above mention hierarchical metadata output in XML. The metadata output uses some terminologies from the current MPEG-7.
|Cite as: Yao, A. and Jin, J.S. (2001). The Development of a Video Metadata Authoring and Browsing System in XML. In Proc. Selected papers from Pan-Sydney Area Workshop on Visual Information Processing (VIP2000), Sydney, Australia. CRPIT, 2. Eades, P. and Jin, J., Eds. ACS. 39-46. |
(local if available)