Towards Adaptive and Resource-Efficient Live Volumetric Video Streaming

Volumetric video enables fully immersive six-degrees-of-freedom (6DoF) experiences. Still, its adoption in real-time applications remains limited by extreme bandwidth demands, high computational cost, and stringent end-to-end latency requirements. This thesis investigates these challenges and introduces three complementary systems—LiV, Super-VV, and RapidX—that collectively advance the practicality, scalability, and resource efficiency of live volumetric video streaming. We first introduce LiV, a server-side adaptive streaming framework integrating viewport-aware tiling, encoding-time prediction, and a Mixed-Integer Linear Programming (MILP)-based quality selector within a Dynamic Adaptive Streaming over HTTP (DASH) architecture. By jointly considering network dynamics, computational constraints, and the user viewpoint, LiV delivers stall-free, high-quality live volumetric video while achieving substantial bitrate savings without compromising interactivity. Next, we present Super-VV, a lightweight, end-to-end pipeline that combines quality-preserving downsampling, tile-parallel Draco compression, and a central processing unit (CPU)-based super-resolution upsampler to significantly reduce bitrate while maintaining perceptual fidelity. Through a pipelined and in-memory architecture, Super-VV achieves real-time throughput on commodity hardware and reduces bandwidth consumption by up to 98% while preserving high visual quality. We then introduce RapidX, a modular testbed for constructing and evaluating complete capture-to-render pipelines under realistic compute and network conditions. RapidX simplifies system prototyping through containerized workflows and enables controlled evaluation of volumetric streaming systems across diverse scenarios. Together, these contributions demonstrate that adaptive, resource-efficient live volumetric video streaming is achievable through coordinated optimization across compression, adaptive delivery, and system-level orchestration. This thesis moves volumetric streaming closer to widespread deployment in telepresence, augmented reality (AR), virtual reality (VR), and immersive communication applications.