Video coding and transmission for multimedia communications using a 3-D graphics model.

摘要

Multimedia communications include compressing and transmitting enormous amount of multimedia data such as text, voice, video, and animated graphics. Because live video streams occupy the largest bandwidth, very low bit rate video coding becomes key in the development of multimedia communications over existing communication channels such as the telephone line, Internet, and wireless systems.; This dissertation investigates video aspects of networked multimedia communications. For compression, we develop a video coding framework using a 3-D graphics model. For transmission, we prototype video over IP in a desktop Internet videoconferencing system. We integrate networking and coding for video distribution.; From a research perspective, we address core modeling issues such as (1) a three-level signal representation model that integrates video and animated graphics; (2) a nonrigid motion synthesis model simulating facial movements; and (3) a 3-D video analysis-by-synthesis feedback system design.; From an application perspective, we apply computer vision, graphics, networking, and numerical simulation in addition to digital signal processing and conventional video coding to (1) conduct 2-D shape feature analysis from video, (2) map the speaker texture over a 3-D head-and-shoulder graphics model, (3) synthesize graphics and extract an animation script from video, (4) code model parameters and video texture for robustness to packet loss during transmission, and (5) transmit video over the Internet and control the bitrate using the state of the network and receivers.