Dynamic Programming Based Smoothing of VBR Video Traffic

The compression of videos using variable-bit-rate encoding produces constant quality video at the expense of significant bit-rate variability. Intelligent traffic smoothing techniques are needed to support efficient network resource utilization. In this paper, we examine a class of bandwidth smoothing algorithms and propose a novel enhancement that uses dynamic programming (DP) to improve upon their performance. Using a long, constant-quality MJPEG encoded video trace, we evaluate the DP-based smoothing algorithm (DPS) and examine its effects on the network, server, and client resources necessary for the transmission of prerecorded videos.