自适应流提出的一个量化问题

中文摘要 AI 辅助

在自适应码率（ABR）流（大多数互联网视频背后的传输技术）中，每个内容被编码为多种码率，形成由多个版本（renditions）构成的编码阶梯。每个客户端播放其网络带宽所能支撑的最高版本。我们证明，选择该阶梯的问题是对带宽分布的标量量化问题。然而，这个量化问题属于一种特殊类型：客户端的逻辑将每个量化区间的再现值固定在区间的左边缘，且失真测度是单侧质量损失而非平方误差。该约束重塑了经典理论。阶梯的质量差距是其平均传输质量与无限多版本所能达到的质量极限之间的距离。对于最优的n个版本的阶梯，差距G_n^*以Θ(1/n)的速率衰减，而非教科书所述的Θ(1/n²)。此外，当n很大时，n·G_n^*→C_w=1/2(∫√(pQ')dB)²，当阶梯的码率遵循密度√(pQ')时达到该极限，这替代了Panter-Dite的p^(1/3)；其中p为带宽密度，Q为内容的质量-码率曲线。该规律产生了闭式设计规则：分位数梯级放置、达到容差ε所需的版本数n(ε)≈C_w/ε，以及当一个版本的运营成本为κ且质量差距的定价为λ时的经济阶梯规模n^*=√(λC_w/κ)。相同的分析可扩展至采用不同视频分辨率、编解码器、感知质量指标以及现代网页播放器的二维自适应逻辑的阶梯设计。

英文摘要

In adaptive bitrate (ABR) streaming, the delivery technology behind most Internet video, each title is encoded at several bitrates, forming an \emph{encoding ladder} of \emph{renditions}. Each client plays the highest rendition that its network bandwidth can sustain. We show that choosing the ladder is a problem of \emph{scalar quantization} of the bandwidth distribution. However, this quantization problem is of an unusual kind: the client's logic pins each quantization cell's reproduction value to the cell's \emph{left edge}, and the distortion measure is a one-sided quality loss rather than a squared error. The constraint reshapes the classical theory. The \emph{quality gap} of a ladder is the distance between its average delivered quality and the \emph{quality limit} that infinitely many renditions would attain. For optimal $n$-rendition ladders the gap $\mathcal{G}_n^*$ decays as $Θ(1/n)$, not the textbook $Θ(1/n^2)$. Moreover, with large $n$, $n\,\mathcal{G}_n^*\to C_w=\frac{1}{2}(\int\!\sqrt{p\,Q'}\,\,d B)^2$, with the limit attained when the ladder's rates follow the density $\sqrt{p\,Q'}$, which replaces Panter--Dite's $p^{1/3}$; here $p$ is the bandwidth density and $Q$ the quality--rate curve of the content. The law yields closed-form design rules: quantile rung placement, the rendition count $n(\varepsilon)\approx C_w/\varepsilon$ needed to reach tolerance $\varepsilon$, and the economic ladder size $n^*=\sqrt{λC_w/κ}$ when a rendition costs $κ$ to operate and the quality gap is priced at $λ$. The same analysis extends to the design of ladders employing different video resolutions, codecs, perceptual quality metrics, and the two-dimensional adaptation logic of modern web players.

发表机构

• Massachusetts Institute of Technology（麻省理工学院）

机构由 AI 辅助整理，请以论文原文为准。