An algorithm for lossless smoothing of MPEG video

Interframe compression techniques, such as those used in MPEG video, give rise to a coded bit stream where picture sizes differ by a factor of 10 or more. As a result, buffering is needed to reduce (smooth) rate fluctuations of encoder output from one picture to the next; without smoothing, the performance of networks that carry such video traffic would be adversely affected. Various techniques have been suggested for controlling the output rate of a VBR encoder to alleviate network congestion or prevent buffer overflow. Most of these techniques, however, are lossy, and should be used only as a last resort. In this paper, we design and specify an algorithm for lossless smoothing. The algorithm is characterized by three parameters: D (delay bound), K (number of pictures with known sizes), and H (lookahead interval). We present a theorem which guarantees that, if K ≥ 1, the algorithm finds a solution that satisfies the delay bound. (Although the algorithm and theorem were motivated by MPEG video, they are applicable to the smoothing of compressed video in general). To study performance characteristics of the algorithm, we conducted a large number of experiments using statistics from four MPEG video sequences.