Tiled streaming for layered 3D virtual reality videos with viewport prediction

In recent years, the demand of 3D video services has gradually increased. More and more bandwidth hungry applications are proposed, such as immersive media services which need a virtual reality (VR) headset and 3D VR videos to provide users immersive experience of watching 3D VR videos. Tiled streaming is often used for providing 3D VR videos die to the high bitrates of 3D VR videos. In a VR headset the sight is limited. Users can only watch a fraction of entire 3D VR videos in the viewport. Transmitting the content of a VR video outside the viewport is unnecessary and infeasible due to the high bitarate of 3D VR videos. Generally, In VR applications, content within the viewport should keep the highest quality while providing only basic quality outside the viewprot. Such an adaptive streaming relies on the precision of viewport prediction. Errors of viewport prediction result in the expensive overhead of quality repairing and re-transmission delay on the traditional versioned coded VR videos,, where a whole new version of video content needs to be resent and the sent content of low-quality version is discarded and cannot be reused. In this paper, we propose a novel adaptative streaming approach for providing 3D VR videos using Scalable Video Coding (SVC) with viewport prediction. For better quality adaptation, we take CubeMap projection as the projection format of 3D VR videos. Besides, we use the Scalability extension of High Efficiency Video Coding (SHVC) to encode 3D VR videos to multiple layers for supplying different qualities and also use the tiling to divide videos into rectangular regions for finer quality adaptation. The experimental results show that our proposed method outperformed other previous approaches in terms of the weighted video quality and the relative time spent on the highest quality, especially with low available network bandwidth. Even under certain miss rates of tiles, compared to previous approaches, our proposed method requires fewer bandwidth overhead and shorter re-transmission delay for repairing the quality of missed tiles in most cases.