12/01/1996 09:00 AM
Computer Science
Digital video coding exploits redundancy both within and between images to achieve the high levels of compression required to support video transmission over dedicated wired links. Video protocols that operate over packet networks like the Internet must additionally take into account the potential for congestion-based packet losses, high latency, and variable bandwidth. Wireless video protocols must contend with even higher packet loss due to uncorrelated phenomena. Unfortunately, these channel characteristics interfere with typical codecs, exacerbating the challenge of maintaining frame rate during periods of moderate packet loss. New codecs that mitigate packet loss by organizing intraframe data in progressive or layered representations are needed for robust wireless video transport; additional compression gains alone are unlikely to provide robustness when packets are lost or corrupted. In addition, new network protocols that allow corrupted packets to reach video applications are required to significantly improve codec robustness beyond current levels. A new metric which evaluates video codestreams on their ability to withstand packet errors is developed and applied to several image transformations, including the DCT and DPCM. Use of the metric suggests a new video codec that is less susceptible to corruption than conventional techniques. Implementation and testing of this SlugVideo codec provides further evidence that current network protocols are inadequate to the task of supporting robust video transmission over lossy packet networks.
