PrismGS: Physically-Grounded Anti-Aliasing for High-Fidelity Large-Scale 3D Gaussian Splatting

TL;DR

Practical problem: aliasing and instability hinder real-time, high-fidelity rendering of large-scale urban scenes with 3D Gaussian Splatting. The authors propose PrismGS, a physically-grounded regularization framework that injects cross-scale anti-aliasing and geometric stability directly into Gaussian primitives via two regularizers: pyramidal multi-scale supervision and Gaussian size regularization. The method operates within a scalable block-based training pipeline and achieves state-of-the-art performance on MatrixCity, Mill-19, and UrbanScene3D, including robust 4K rendering with notable PSNR and perceptual gains. This work reduces flickering textures and jagged edges, enabling more reliable digital twin and urban visualization applications, with PSNR gains around 1.0–1.5 dB against CityGaussian.

Abstract

3D Gaussian Splatting (3DGS) has recently enabled real-time photorealistic rendering in compact scenes, but scaling to large urban environments introduces severe aliasing artifacts and optimization instability, especially under high-resolution (e.g., 4K) rendering. These artifacts, manifesting as flickering textures and jagged edges, arise from the mismatch between Gaussian primitives and the multi-scale nature of urban geometry. While existing ``divide-and-conquer'' pipelines address scalability, they fail to resolve this fidelity gap. In this paper, we propose PrismGS, a physically-grounded regularization framework that improves the intrinsic rendering behavior of 3D Gaussians. PrismGS integrates two synergistic regularizers. The first is pyramidal multi-scale supervision, which enforces consistency by supervising the rendering against a pre-filtered image pyramid. This compels the model to learn an inherently anti-aliased representation that remains coherent across different viewing scales, directly mitigating flickering textures. This is complemented by an explicit size regularization that imposes a physically-grounded lower bound on the dimensions of the 3D Gaussians. This prevents the formation of degenerate, view-dependent primitives, leading to more stable and plausible geometric surfaces and reducing jagged edges. Our method is plug-and-play and compatible with existing pipelines. Extensive experiments on MatrixCity, Mill-19, and UrbanScene3D demonstrate that PrismGS achieves state-of-the-art performance, yielding significant PSNR gains around 1.5 dB against CityGaussian, while maintaining its superior quality and robustness under demanding 4K rendering.

Figures (3)

• Figure 1: Overview of the PrismGS framework. Our method first partitions the scene into blocks for parallel training. During optimization, we introduce two key regularizers: a Multi-Scale Supervision loss for anti-aliasing and a Size Regularization loss for geometric stability.
• Figure 2: Qualitative comparisons of different methods (Mega-NeRF, Octree-GS, CityGaussian, Momentum-GS, Ours) against Ground Truth across four large-scale scenes. Orange insets highlight patches that reveal notable visual differences, demonstrating the superiority of our method in capturing fine details and maintaining structural consistency.
• Figure 3: Qualitative results of ablation study. Excluding any module leads to lower reconstruction quality and other impacts.