Smooth Viewport Bitrate Adaptation for 360 Video Streaming

Abstract

In Virtual Reality, 360-degree video is a major content type that provides immersive viewing experience for users. In this paper, we propose S-VAS, a new adaptation method for 360-degree video streaming, which can provide users with smooth viewport bitrate under network bandwidth variations and time-varying user head movements. Our proposed method adapts the number of extension tiles according to viewport prediction errors. Besides, variations in bitrate between adjacent tiles in a segment are mitigated by a novel tile version selection algorithm. Experimental results show that the proposed method can reduce the inter-segment bitrate variation by up to 46% and the intra-segment bitrate variation by up to 32%.