Spatial scalable video service has surged in the time of multiple screens. Existing bitrate allocation methods are principled by rate-distortion theory and characterized by iterative encoding, which is accurate yet complex. However, the quasi-quantitative description is preferred in practice of broadcasting. In this paper, we propose a task of bitrate estimation for scalable videos concerning the content, aiming at a more efficient model at the cost of precision. First, we exhibit necessity to build a model for Scalable High Efficiency Video Coding (SHVC) and quantitative relation between video content and bitrate using different encoders. Then, a scalable-video dataset is prepared. It covers various types of content to offer diversity for model training. In the end, multi-linear regression is utilized to estimate the bitrate of scalable videos, with spatial and temporal indices as explanatory variables. Our statistical experiments show the model is able to estimate bitrate after trained on the self-built dataset.

中文翻译：

---

空间可缩放视频的比特率估计

空间可伸缩视频业务在多屏时代激增。现有的比特率分配方法以率失真理论为原则，以迭代编码为特征，准确而复杂。然而，在广播实践中更倾向于准定量描述。在本文中，我们提出了一个关于内容的可伸缩视频的比特率估计任务，旨在以精度为代价获得更有效的模型。首先，我们展示了使用不同编码器为可扩展高效视频编码 (SHVC) 和视频内容与比特率之间的定量关系构建模型的必要性。然后，准备可伸缩视频数据集。它涵盖了各种类型的内容，为模型训练提供多样性。最后，利用多元线性回归来估计可伸缩视频的比特率，以空间和时间指数作为解释变量。我们的统计实验表明，该模型在自建数据集上训练后能够估计比特率。