Video streaming continues to captivate attention of users and service providers, dominate in Internet traffic, and form a vibrant research field. Taking a pragmatic approach to reviewing recent research in the field, this paper considers the most dominant streaming paradigm, the main aspects of which include transmission of two-dimensional videos over the best-effort Internet, support from content delivery networks, and client-side bitrate adaptation. To make the survey more accessible, we incorporate extensive tutorial materials. In contrast with the siloed approaches of existing surveys, our paper holistically covers the end-to-end streaming pipeline from video capture and upload for server processing to distribution for playback on diverse user devices. Reflecting the practical interests of respective stakeholders, our survey presents a novel perspective on end-to-end streaming and sheds light on the relationships and interactions between its ingestion, processing, and distribution stages. At each stage, we classify streaming designs in regard to their methodology depending on whether intuition, theory, or machine learning serves as a methodological basis for their core contribution. In addition to tasks confined to a single stage, the survey also examines transversal topics such as coding, super resolution, and quality of experience. After surveying more than 200 papers, we synthesize current trends and project future directions in video streaming research.

翻译：视频流持续吸引着用户与服务提供商的关注，主导着互联网流量，并构成一个蓬勃发展的研究领域。本文采用务实方法回顾该领域的最新研究，聚焦于主流的流媒体范式，其主要方面包括：通过尽力而为互联网传输二维视频、内容分发网络的支持以及客户端码率自适应。为增强综述的可读性，我们融入了大量教程材料。与现有综述的孤立视角不同，本文整体覆盖了从视频采集上传至服务器处理，再到分发到各类用户设备播放的端到端流管道。反映各利益相关方的实际需求，我们的综述提出了端到端流媒体分析的新视角，揭示了其摄取、处理与分发阶段之间的关联与交互。在每个阶段，我们根据设计方法的核心贡献基础（直觉、理论或机器学习）对设计方案进行分类。除局限于单个阶段的任务外，本文还深入探讨了编码、超分辨率与体验质量等跨领域主题。在调研200余篇论文的基础上，我们综合了视频流研究的当前趋势，并展望了未来方向。