Scalable Free-Viewpoint 3D Video Streaming

点击翻译按钮获取中文摘要

Traditional video systems present only a single view of the 3D world to viewers. For a more realistic experience; however, one should to be able to view a scene from different viewpoints, as in the real world. While 3D video enables viewers to perceive depth, 3D alone is also limited to a specific viewing angle and does not show different viewpoints. In contrast, free-viewpoint video enables viewers to navigate freely to any viewpoint and allows them to see interesting aspects in the video and enjoy a rich and immersive experience. However, free-viewpoint videos face multiple research challenges that have so far prevented their wide adoption. First, a user may request a viewpoint that was not captured by the cameras. This requires synthesizing, in real-time, the requested view from existing views, which is a computationally expensive process. Second, the streaming server needs to be scalable in order to support many viewers interacting with it. Third, viewers typically use different displays, from smartphones and tablets to large displays, and also use different network access such as 4G, 3G, ADSL, WiMax, WiFi, etc. Therefore, a free-viewpoint system must support clients with heterogeneous requirements. To address the above challenges, in this project we propose to work with Cisco Systems Canada, specifically their Video and Content Platforms Research and Advanced Development group in Toronto, Ontario, Canada, to study, design, and develop a free-viewpoint streaming system. Cisco builds and sells industry-leading video streaming systems, such as their Videoscape platform, which powers subscriber experiences that are personalized, synchronized, and more social than existing products.