The area of content-based video retrieval is a very hot area both for
research and for commercial applications. In order to design effective
video databases for applications such as digital libraries, video
production, and a variety of Internet applications, there is a great
need to develop effective techniques for content-based video retrieval.
One of the main issues in this area of research is how to bridge the
semantic gap between low-Ievel features extracted from a video (such as
color, texture, shape, motion, and others) and semantics that describe
video concept on a higher level. In this book, Dr. Milan Petkovi6 and
Prof. Dr. Willem Jonker have addressed this issue by developing and
describing several innovative techniques to bridge the semantic gap. The
main contribution of their research, which is the core of the book, is
the development of three techniques for bridging the semantic gap: (1) a
technique that uses the spatio-temporal extension of the Cobra
framework, (2) a technique based on hidden Markov models, and (3) a
technique based on Bayesian belief networks. To evaluate performance of
these techniques, the authors have conducted a number of experiments
using real video data. The book also discusses domains solutions versus
general solution of the problem. Petkovi6 and Jonker proposed a solution
that allows a system to be applied in multiple domains with minimal
adjustments. They also designed and described a prototype video database
management system, which is based on techniques they proposed in the
book.
