Efficient viewport prediction and tiling schemes for 360 degree video streaming

Abstract

360-degree video streaming for VR visualisation is characterised by large transmission data volume and stringent interactive latency demands; hence guaranteeing suitable transmission quality, while meeting the existing constraints, is highly challenging. This paper addresses the relevant grand challenge presented at MMSys 2024 on 360-degree video on-demand streaming, aiming at designing and implementing a 360-degree video on-demand streaming solution using the open-source evaluation platform E3PO. The proposed solution incorporates several strategies including viewport prediction, tiling, encoding tile selection for streaming and upsampling. The source code for the proposed solution will be publicly available on Github1.