VIRON: An Annotation-Based Video Information Retrieval System

T o provide e f i c i e n t search and retrieval of video data f r o m large archives, we need t o model video data appropriately. In this paper we propose a video data model an,d describe the design and implementa t ion of the annotation-based video retrieval s y s t e m VIRON (Video I n f o r m a t i o n Retr ieval On Nota t ion) based o n the proposed model . T h i s model provides the mechanism of sharing and reusing annotat ions among users by introducing descriptor schema. In order t o process query e f ic ien t ly , a n annotated video unit i s mapped i n t o a n unified video annota t ion s tream. V I R O N i s composed of three tools: C V U manager , which i s used t o manage and visualize conceptual video units; annota tor t o annota te video data by of fer ing interactive video player; and video query tool to pose and process video queries. video retrieval [3 , 4, 181, and (2) annotation-based video retrieval [6 , 10, 14, 15, 16, 171. Even though the state-of-the-art image processing technologies make it possible to automatically analyze scene breaks and pauses in audio [19, 201, posing and processing video queries are still technically difficult. In this paper we consider the issue of how users could manipulate the video data more efficiently. We propose a video data model based on the video annotation. In our model, it is possible to represent video information as structured data, provided that computer-supported human annotation could support reasonable facilities for annotation and retrieval of video data. 2 Related works and our approach