| In this work, we study the relationships between video quality, bit rate variability, and the utility from a streaming service with statistical multiplexing for open-loop encoded video. We begin with a thorough overview of the rate variability-distortion (VD) curve of open-loop encoded video, which we have found to exhibit a characteristic "hump" behavior. We investigate how this hump behavior is influenced by the different motion levels in the video content, the video encoding parameters, and traffic smoothing.; Evaluating different levels of approximation and their accuracy in capturing the VD curve's characteristics, we find that aggregation levels above the macroblock level tend to decrease the accuracy of approximations.; We describe a methodology for assessing the admissibility of a set of homogenous video streams on a link subject to a statistical quality of service criterion. We extend this network-centric view to a service-provider view which includes the assessment of the utility (revenue) earned by statistically multiplexing video of different quality levels.; To facilitate the evaluation of streaming mechanisms by simulation, we additionally review the shortcomings of the currently available video traces with respect to their ability to facilitate networking research with exact video stream quality evaluation after lossy network transport. Towards this end, we introduce offset distortion video traces that allow networking researchers to accurately calculate the video frame distortion and quality in case of re-displayed frames. |
