A Video Retrieval and Sequencing System

Abstract

Video is an effective medium for capturing the events in the real world around us, and a vast amount of video materials exists, covering a wide range of applications. However, widespread use of video in computer applications is often impeded by the lack of effective tools to manage video information systematically. This article discusses the design and implementation of a frame-based video retrieval and sequencing system (VRSS). The system is designed to support the entire process of video information management: segmenting, indexing, retrieving, and sequencing of video data. A semiautomatic tool is developed to divide video sequences into meaningful shots, Each video shot is logged using text descriptions, audio dialogue, and cinematic attributes. A two-layered, concept-based model is used as the basis for accurately retrieving relevant video shots based on users' free-text queries. A cinematic, rule-based, virtual editing tool is also developed to sequence the video shots retrieved for presentation within a specified time constraint. The system has been tested on a video documentary on the NUS (National University of Singapore) engineering faculty. The results of video retrieval experiments are encouraging. Categories and Subject Descriptors: H.3.O [Information Storage and Retrieval]: General; H.5. 1 [Information Interfaces and Presentation]: Multimedia Information Systems; 1.2.4 [Artificial Intelligence]: Knowledge Representation Formalisms and Methods General Terms: Experimentation. © 1995, ACM. All rights reserved.