Gaussian Splatting-based Volumetric Video Compression with Sparse 4D Anchors

2026-09-27

AI Summary

This work addresses the compression challenges in dynamic 3D Gaussian Splatting caused by dense primitives and spatiotemporal redundancy, as well as the limitations of existing anchor-based methods in handling non-local dynamics and occlusions. To this end, it proposes SAGA, a volumetric video codec that represents dynamic scenes using hierarchical sparse 4D anchors. A coordinate-based implicit neural representation (INR) decoder generates fine-grained anchors and Gaussian primitives to enable compact parameter sharing. Furthermore, fixed-size memory slots with an orthogonality update mechanism are introduced to effectively model long-range dependencies among unstructured anchors, substantially improving entropy context modeling accuracy. Evaluated on the Neu3D and MPEG MIV datasets, SAGA achieves PSNR BD-rate reductions of 80.39% and 83.94%, respectively, compared to GIFStream, demonstrating superior rate-distortion performance.

Abstract

Immersive video communication requires photorealistic, render-efficient, and compact dynamic scene representations. 3D Gaussian Splatting (3DGS) offers a promising representation, but dynamic 3DGS remains difficult to compress due to dense primitives and spatiotemporal redundancy. Anchor-based formulations improve compactness with sparse scaffolds that share geometry and appearance across primitives. However, existing designs often rely on deforming a single canonical scaffold and condition each primitive on its associated anchor in isolation, limiting their ability to handle non-local dynamics and disocclusion while under-exploiting inter-anchor correlations, particularly in motion- or texture-dense regions. To address these limitations, we propose SAGA, a volumetric video codec built upon Sparse Anchor-assisted GAussian splatting representations. SAGA represents dynamic 3D scenes using hierarchically organized sparse 4D anchors, where coordinate-based INR decoders generate fine anchors and Gaussian primitives from inter-anchor interpolations, enabling compact parameter sharing across spatiotemporal structures. For long-range dependencies among unstructured anchors, we further introduce fixed-size memory slots with orthogonality-informed updates for accurate entropy-context modeling. Experiments show that SAGA achieves strong rate-distortion performance against GIFStream, with PSNR BD-rate reductions of 80.39% and 83.94% on Neu3D and MPEG MIV, respectively.

Problem

• Volumetric Video Compression
• 3D Gaussian Splatting
• Dynamic Scene Representation
• Spatiotemporal Redundancy
• Anchor-based Formulation

Innovation

• Sparse 4D Anchors
• 3D Gaussian Splatting
• Volumetric Video Compression
• INR Decoders
• Entropy-context Modeling