Eye Fixation and Saliency Field based Greedy Method for Volumetric Video Streaming

Metaverse megatrend sparks interest in volumetric videos, especially point cloud data format. However, point cloud videos usually contain large amounts of data, which makes it quite difficult to achieve efficient transmission under limited bandwidth. In this paper, we propose a greedy method named GES tailored for volumetric video streaming, which exploits human eye fixation data and saliency field to allocate bitrates for different tiles decomposed from point cloud scenes under a given bandwidth constraint. Moreover, to evaluate the performance of GES, we construct a platform to collect user head posture and eye fixation data, accumulating 40 user viewing behavior trajectories in total. Extensive experiments illustrate that our proposed method achieves superior performance than existing representative methods on real user behavior data, especially in visual performance.