Advancing User Quality of Experience Using Viewport Archives in Viewport-Aware Tile-Based 360-Degree Video Streaming

In tile-based 360-degree video streaming, the video content is temporally divided into segments, wherein the content is further divided spatially into multiple partitions called tiles. As a common technique to reduce the data size for the segment in tile-based streaming, various viewport prediction models are adopted to allocate bandwidth to the tiles within the region predicted for the user, which is also called the quality enhanced region or viewport. However, due to the unpredictable nature of user head movements, it is difficult to accurately estimate the viewport, resulting in great degradation in the Quality of Experience (QoE) for the user. In this paper, to improve user QoE for even when there are estimation errors, we propose extending the viewport to include the direction(s) holding the highest likelihood to be seen by the user by assigning weights to each tile of each segment from the viewport archives of past users. From our performance evaluation on two 360-degree videos, the proposed method demonstrated a significant increase in the perceived viewport quality while suppressing quality variation within the viewport and across segments compared to two reference models.