GS^2: Graph-based Spatial Distribution Optimization for Compact 3D Gaussian Splatting

2026-04-02

AI Summary

This work addresses the high memory overhead of 3D Gaussian Splatting (3DGS) and the tendency of existing pruning methods to disrupt spatial coherence and introduce rendering artifacts. The authors propose a compact yet high-fidelity optimization framework that jointly refines Gaussian distributions through three core strategies: ELBO-based adaptive densification, opacity-aware progressive pruning, and graph neural network–driven feature encoding coupled with positional redistribution. By synergistically integrating these components, the method achieves superior rendering quality—exceeding the original 3DGS in PSNR—while utilizing only approximately 12.5% of the original Gaussian count, thereby substantially reducing memory consumption and outperforming current state-of-the-art baselines.

Abstract

3D Gaussian Splatting (3DGS) has demonstrated breakthrough performance in novel view synthesis and real-time rendering. Nevertheless, its practicality is constrained by the high memory cost due to a huge number of Gaussian points. Many pruning-based 3DGS variants have been proposed for memory saving, but often compromise spatial consistency and may lead to rendering artifacts. To address this issue, we propose graph-based spatial distribution optimization for compact 3D Gaussian Splatting (GS\textasciicircum2), which enhances reconstruction quality by optimizing the spatial distribution of Gaussian points. Specifically, we introduce an evidence lower bound (ELBO)-based adaptive densification strategy that automatically controls the densification process. In addition, an opacity-aware progressive pruning strategy is proposed to further reduce memory consumption by dynamically removing low-opacity Gaussian points. Furthermore, we propose a graph-based feature encoding module to adjust the spatial distribution via feature-guided point shifting. Extensive experiments validate that GS\textasciicircum2 achieves a compact Gaussian representation while delivering superior rendering quality. Compared with 3DGS, it achieves higher PSNR with only about 12.5\% Gaussian points. Furthermore, it outperforms all compared baselines in both rendering quality and memory efficiency.

Problem

• 3D Gaussian Splatting
• memory efficiency
• spatial consistency
• rendering artifacts
• compact representation

Innovation

• Graph-based optimization
• 3D Gaussian Splatting
• Spatial distribution
• Adaptive densification
• Opacity-aware pruning