Performance evaluation of projection and compression in 3D 360 video

3D 360 视频中投影和压缩的性能评估

• 批准号: 523516-2018
• 负责人: Coulombe, Stéphane
• 金额: $ 1.82万
• 依托单位: École de technologie supérieure
• 依托单位国家: 加拿大
• 项目类别: Engage Grants Program
• 财政年份: 2018
• 资助国家: 加拿大
• 起止时间: 2018-01-01 至 2019-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=648193
• 关键词: Performance; evaluation; projection; compression; 3D

Omnidirectional (360) video applications are diverse and increasingly popular. They include immersive films,**live-streamed concerts, gaming, tourism, shopping and telepresence. Due to human peripheral vision being**limited to a small area of the 360 videos, and the area in view must be of sufficiently high-enough resolution, it**forces the capture of 360 videos at minimum of 4K or even 8K resolutions. Such high-resolution videos lead to**high computational complexity and huge data rates. These problems are aggravated with stereo (3D) 360**videos which are highly desirable to deliver a true immersive experience. Summit Tech offers 360 video**solutions and faces challenges with current streaming capabilities within networks constrained in delivering**optimal performance due to changing network conditions. It needs to assess which projections and encoders**can provide the best compression/ quality/ computational complexity trade-offs. Moreover, Summit Tech plans**to offer 3D 360 video solutions. Therefore, it needs to determine if the projections that are successful on 360**monoscopic video still perform well on 3D 360 videos. It must also determine if 3D codecs, which exploit the**disparity between views, can perform significantly better than the monoscopic ones on 3D content. The main**objective of this project is to develop software tools that permit to evaluate the performance of various 3D 360**components (e.g. projections and encoders) under various quality metrics. The projections will include**well-known projections as well as newer ones from the literature. The quality metrics will include PSNR-based**as well as novel panoramic and 3D objective quality metrics from the literature. The developed software will**be an extension of MPEG's JVET 360Lib supporting a wider range of projections, encoders and quality**metrics. The proposed software will also be reengineered to support the processing and performance evaluation**of 3D 360 videos which is currently lacking in 360Lib. This new software tool, the first supporting 3D 360**videos, will provide crucial information to orient the strategic technical directions and designs of new and**improved Summit Tech products and ensure their continuous world leadership and success.

全方位（360）视频应用是多种多样的，并且日益流行。它们包括沉浸式电影、直播音乐会、游戏、旅游、购物和远程呈现。由于人的周边视觉**局限于360视频的一小块区域，而所看到的区域必须有足够高的分辨率，这**迫使360视频的捕获至少4K甚至8K的分辨率。如此高分辨率的视频导致高计算复杂度和巨大的数据速率。这些问题在立体（3D） 360度视频中更加严重，而立体（3D） 360度视频非常适合提供真正的沉浸式体验。Summit Tech提供360视频解决方案，并面临当前网络流媒体能力的挑战，由于网络条件的变化，网络在提供最佳性能方面受到限制。它需要评估哪种投影和编码器**可以提供最佳的压缩/质量/计算复杂性权衡。此外，顶峰科技计划**提供3D 360视频解决方案。因此，需要确定在360**单镜视频上成功的投影在3D 360视频上是否仍然表现良好。它还必须确定3D编解码器（利用视图之间的差异）在3D内容上的表现是否明显优于单视角编解码器。该项目的主要目标是开发软件工具，允许在各种质量指标下评估各种3D 360**组件（例如投影和编码器）的性能。这些预测将包括著名的预测以及来自文献的新预测。质量指标将包括基于psnr的**以及来自文献的新颖全景和3D客观质量指标。开发的软件将是MPEG的JVET 360Lib的扩展，支持更广泛的投影，编码器和质量指标。提议的软件也将被重新设计，以支持3D 360视频的处理和性能评估**，这是目前360Lib所缺乏的。这个新的软件工具，第一个支持3D 360视频，将提供关键信息，以确定新的和改进的顶峰科技产品的战略技术方向和设计，并确保他们持续的世界领先地位和成功。