A control-theoretic approach to adapting VBR compressed video for transport over a CBR communications channel

Future broad-band communications networks are expected to be dominated by video and image traffic. Variable bit-rate (VBR) video compression is generally preferred to constant bit-rate (CBR) compression because constant image quality can be provided. In contrast, CBR transport is preferred to VBR transport from the networking standpoint because of its simplicity. This paper studies the important issue of adapting VBR compressed video for transport over a CBR channel. We focus on temporal traffic smoothing using an elastic buffer. The target image quality and the output rate of the video encoder is controlled by feedback based on the buffer-occupancy level. Previous adaptation schemes are not readily analyzable. An analyzable control-theoretic adaptation framework is proposed. It allows systematic and quantitative investigation of issues such as stability, robustness against scene changes, robustness against image-quality oscillations due to coding-mode switching, and tradeoffs between image-quality and buffer-occupancy (delay) fluctuations. Perhaps more importantly, the framework opens up many new possibilities for further research.