Upper limits of head orientation prediction for 360-degree video streaming

Delivery of 360-degree videos poses a challenge to the existing video distribution chains due to the high volume of data to be transmitted. Viewport-aware delivery schemes, e.g. tilebased streaming, constitute a promising approach to reduce the transmitted data volume but require knowledge of the future user viewport to reach its full potential. We discuss a predictive tile-based streaming scenario using head orientation prediction and study the upper limits in terms of throughput savings for different anticipation times and system end-to-end delays. Our results show that using head motion prediction for tile-based streaming systems can bring throughput gains up to 46% compared to a benchmark tile-based streaming system without prediction.