Lightweight 3D Animation Rendering for Real‐Time Immersive Communication

ABSTRACT Immersive 3D communication is an emerging paradigm in the Next‐Generation Internet (NGI) and the Metaverse, enabling volumetric telepresence beyond conventional 2D video streaming. While systems like VirtualCube have demonstrated high‐fidelity real‐time telepresence, they rely on computationally intensive Multi‐View Stereo (MVS) pipelines that impose severe latency penalties and require workstation‐class hardware, making them unsuitable for resource‐constrained edge devices. To address the critical bottleneck of end‐to‐end latency in real‐time immersive services, this paper proposes Neural Implicit Surface Anchor (NISA) Rendering, a lightweight, edge‐friendly framework whose core philosophy is to shift expensive per‐pixel processing to a sparse, semantic‐aware paradigm. This is realized through two complementary contributions: (1) a sparse‐to‐dense geometric anchor mechanism that replaces dense per‐pixel depth transmission with a minimal set of semantic keypoints, targeting geometric efficiency; and (2) a silhouette‐gated refinement network that skips redundant computations in flat regions, targeting pixel‐level computation efficiency. Experimental results demonstrate that our method reduces floating point operations (FLOPs) by 65% compared to baseline volumetric systems, achieving stable 3D rendering at 52 FPS on consumer‐grade GPUs with a PSNR of 30.15 dB. This approach effectively balances high quality of experience (QoE) with the low‐latency constraints of modern internet infrastructure.