Olejar Books

This book focuses particularly on content-based video retrieval. After addressing basic concepts and techniques in the field, Content-Based Video Retrieval: A Database Perspective concentrates on the semantic gap problem, i.e., the problem of inferring semantics from raw video data, as the main problem of content-based video retrieval. This book identifies and proposes the integrated use of three different techniques to bridge the semantic gap, namely, spatio-temporal formalization methods, hidden Markov models, and dynamic Bayesian networks. As the problem is approached from a database perspective, the emphasis evolves from a database management system into a video database management system. This system allows a user to retrieve the desired video sequence among voluminous amounts of video data in an efficient and semantically meaningful way. This book also presents a modeling framework and a prototype of a content-based video management system that integrates the three methods and provides efficient, flexible, and scalable content-based video retrieval. The proposed approach is validated in the domain of sport videos for which some experimental results are presented.