Towards Practical and Resource-Efficient Volumetric Video Streaming

Abstract

Volumetric video offers immersive, six-degrees-of-freedom (6DoF) experiences but presents significant challenges in streaming due to its massive data size, high bitrate demands, and computational complexity. This thesis proposes two systems to address these challenges. First, we present VV-DASH, an end-to-end framework for adaptive volumetric video streaming over DASH. VV-DASH introduces a codec-agnostic segment format (DVV) that consolidates compressed volumetric content into DASH-ready segments, improving streaming throughput by 13.2% and reducing bandwidth demands. It also features a parallel accelerator that enables real-time decoding—achieving up to 183 fps for Draco and 30 fps for V-PCC—while supporting flexible integration with diverse codec pipelines. Next, we introduce TARS, a temporal-spatial adaptive streaming solution for dynamic point cloud videos that reduces temporal redundancy by exploiting inter-frame correlations. TARS employs a Field-of-View (FoV)-aware approach to intelligently avoid the retransmission of redundant regions across consecutive frames, leveraging a specialized Point Cloud Structural Similarity metric for precise similarity assessment. Experiments show that TARS reduces bandwidth usage by up to 66% while maintaining high visual quality (MSE as low as 0.145), and boosts decoding speed by up to 2.42x compared to decoding all I-frames. These results demonstrate the effectiveness of VV-DASH and TARS for enabling practical, real-time, and resource-efficient volumetric video streaming.