Collaborative Research: CNS Core: Small: From Capture to Consumption: System Challenges in Pervasive 360-Degree Video Sharing

360度视频允许用户从相机位置的任何角度探索录制的场景。360度视频格式提供的这种更大的灵活性使其越来越受欢迎。但是，由于这种格式需要大量的资源，广泛采用的速度放慢了。这项工作提出了在360度视频共享路径的每个阶段提高效率的方法，从发送端的视频捕获到接收端的消费。该项目通过四个研究重点来解决这些挑战。第一个推力探讨了在使用商品镜头捕获的多个视频实时拼接期间生成高质量帧的方法。本文提出的方法采用梯度变换来细化初始粗糙拼接。第二部分研究基于可用带宽对上传内容进行空间调整的机制。第三个目标是优化全向内容的细粒度表示，以达到较高的投影效率和观看效率。最后一个推力使用边缘计算技术来优化接收端预审窗口大小，以允许带宽高效的内容下载。在项目过程中开发的技术将允许用户享受更高效的360度视频流应用程序。这些技术不仅适用于360度视频流，而且将在未来几代虚拟现实（VR）技术中得到应用。该项目还将激发一般学生对计算机科学研究的兴趣，在项目研究过程中直接训练学生，并有助于课程开发。该项目将产生出版物、代码、数据和其他研究工件。所有这些工件将在项目过程中通过url http://www.cs.gmu.edu/~sqchen/和http://www.cs.binghamton.edu/~yaoliu/公开提供，并在项目完成后至少五年内保持可用。代码将是开源的，也将在代码托管网站，如GitHub上提供。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

360-degree video allows users to explore a recorded scene at any angle from the camera position. This greater flexibility provided by the 360-degree video format has led to its increased popularity. However, widespread adoption has slowed because of the large amounts of resources required by the format. This work proposes methods to improve efficiency at each stage of the 360-degree video sharing path, from video capture at the sender side to consumption at the receiver side. The project addresses these challenges via four research thrusts. The first thrust explores approaches for generating high-quality frames during real-time stitching of multiple videos captured using commodity lenses. The proposed approach applies gradient decent to refine an initial rough stitching. The second thrust investigates mechanisms to spatially adapt content to be uploaded based on the available bandwidth. The third thrust aims to optimize the fine-grained representation of omnidirectional content to achieve high projection efficiency and view efficiency. The last thrust uses edge computing techniques to optimize the look-ahead window size on the receiver side to allow bandwidth-efficient content downloading.The techniques developed during the course of the project will allow users to enjoy more efficient 360-degree video streaming applications. These techniques are not only applicable to 360-degree video streaming but will also have applications in future generations of virtual reality (VR) technologies. The project will also motivate general student interest in computer science research, directly train students during the course of the project’s research, and contribute to curriculum development. The project will produce publications, code, data, and other research artifacts. All such artifacts will be made available publicly through the URLs http://www.cs.gmu.edu/~sqchen/ and http://www.cs.binghamton.edu/~yaoliu/ during the course of the project and remain available for at least five years after completion of the project. The code will be open sourced and will also be made available at code hosting website, e.g., GitHub.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.