Design and evaluation of tile selection algorithms for tiled HTTP adaptive streaming

Abstract

The future of digital video is envisioned to have an increase in both resolution and interactivity. New resolutions like 8k UHDTV are up to 16 times as big in number of pixels compared to current HD video. Interactivity includes the possibility to zoom and pan around in video. We examine Tiled HTTP Adaptive Streaming (TAS) as a technique for supporting these trends and allowing them to be implemented on conventional Internet infrastructure. In this article, we propose three tile selection algorithms, for different use cases (e.g., zooming, panning). A performance evaluation of these algorithms on a TAS testbed, shows that they lead to better bandwidth utilization, higher static Region of Interest (ROI) video quality and higher video quality while manipulating the ROI. We show that we can transmit video at resolutions up to four times larger than existing algorithms during bandwidth drops, which results in a higher quality viewing experience. We can also increase the video quality by up to 40 percent in interactive video, during panning or zooming. © 2013 IFIP International Federation for Information Processing.