Videoconferencing and Videotelephony: Technology and Standards, Second Edition
A concise and complete introduction to the design and deployment of standards-based videoconferencing and videotelephony applications.
TABLE OF CONTENTS Videoconferencing and Videotelephony: Technology and Standards, Second Edition
3.9: Object-Based Coding Techniques
3.9 Object-Based Coding Techniques
One difficulty with the H.261 coding algorithm is that blocking artifacts and mosquito noise are frequently generated at the boundary between moving objects and the static background. This is caused by the fact that the DCT block edges are not aligned with the edge of the object. Object-based coding (OBC) has been devised to attempt to minimize distortions of this type. The general principle of OBC [16 23] involves the identification of, and the encoding of, arbitrarily shaped objects moving within the scene. Information is transmitted about the object defining its motion, shape, and color. Two types of OBC systems have been investigated: a generic approach dealing with unknown objects, and approaches dealing with known objects such as a talking head (also known as knowledge-based coding). The status of work in these two areas is outlined as follows.
3.9.1 Generic Unknown Objects
The University of Hanover, in Germany, has been a leader in the research of OBC. Scholars there have published many articles on the subject and have focused most of their energies on a specific implementation known as object-basedanalysis-synthesis coding (OBASC). Figure 3.19 is a functional block diagram of the OBASC encoder. The general architecture is the same as the H.261 coder in that they both employ interframe prediction and use the classic predictive loop structure. The parameter coder compares the current parameters with the storedand predicted parameters and transmits the prediction errors to the receiver. The loop is a little more complex than normal because...
Copyright Artech House, Inc. 1999 under license agreement with Books24x7
