VAS360: QoE-Driven Viewport Adaptive Streaming for 360 Video
Yuxiang Hu, Yu Liu, Yumei Wang · 2019
Since 360 video requires extremely high bitrate, viewport adaptive streaming is an effective way to transmit 360 video, mainly considering the accurate prefetching of the viewport content and playback fluency. In this paper, we propose VAS360, a 360 video streaming system. In viewport adaptation, we prefetch tiles in advance based on viewport prediction. The innovative tile scheduler downloads compensation tiles based on prediction errors in different directions, which can tolerate inaccuracy in viewport prediction. In QoE-driven rate adaptation, we use the weighted sum of QoE metrics as a utility, maximizing the utility under the constraint of total bitrate. At the same time, our buffer-based rate control can significantly reduce the stall duration. Finally, through extensive evaluation of the experiment, VAS360 does improve the user experience. Compared to non-viewport adaptive schemes, VAS360 increases the viewport quality by 23%-45%.