Multi layer video object database based on interactive annotation and its application

As the use of applications employing video contents becomes widespread, it is crucial for representation systems to be able to handle video effectively and flexibly. In this paper, the authors propose a method that has enriched video contents and an application for future interactive TV service, called Advanced Database TV (ADTV). The system requires three basic methods: one is a video handling method based on video objects; another is a construction method of a video object database achieved by feature extraction and tracking of objects; and the third is a protocol for delivery annotations of objects which are given by users. The objective is to provide a multi-layer video object database based on users' interactive description of video objects. The database will have the ability to organize incomplete descriptions of video objects in each frame, mainly annotation, position and motion. We have implemented the prototype system, which provides users functions such as semi-automatic enrichment of annotations, similarity retrieval of objects, and semantic indexing, based on the video object database.