Abstract
Free-viewpoint video (FVV) applications enable viewers to interactively change their viewing point and watch a scene from different angles. Each FVV is composed of multiple streams representing the scene and its geometry from different vantage points. In addition, virtual views can be synthesized from captured views to provide a smoother and more immersive experience to users. Delivering FVV streaming services over cellular networks, while achieving quality-of-experience (QoE) fairness and minimizing fluctuations in perceived quality, is very challenging due to the large bandwidth requirements, the complex relationship between the bitrates of the transmitted streams and the quality of rendered virtual views, and the time-varying channel conditions. In this paper, we formulate FVV adaptive streaming as a multi-objective QoE-fairness problem and propose a heuristic algorithm to solve it efficiently. Our experiments show that the proposed algorithm achieves high QoE-fairness and provide users with high and stable qualities. It reduces quality variations by up to 32 % on average while saving up to 18 % of cellular bandwidth, compared to state-of-the-art approaches.