Structure-Aware Gaussian Splatting for Large-Scale Scene Reconstruction

2026-07-02

AI Summary

This work addresses the challenge of sparse observations in large-scale Gaussian splatting, which often leads to excessive densification and redundancy, degrading both reconstruction quality and computational efficiency. To mitigate this, the authors propose a structure-aware training framework grounded in signal structure recovery. Central to this approach is the SIG scheduler, which, for the first time, integrates the scene’s frequency convergence behavior into the training dynamics, enabling adaptive coordination between image resolution and Gaussian densification. Additionally, spherical-constrained Gaussians are introduced to incorporate spatial priors, facilitating geometry-aware, float-free optimization with consistent spectral characteristics. Experiments demonstrate that the proposed method significantly outperforms existing techniques in large-scene reconstruction, achieving state-of-the-art performance in both rendering fidelity and computational efficiency.

Abstract

3D Gaussian Splatting has demonstrated remarkable potential in novel view synthesis. In contrast to small-scale scenes, large-scale scenes inevitably contain sparsely observed regions with excessively sparse initial points. In this case, supervising Gaussians initialized from low-frequency sparse points with high-frequency images often induces uncontrolled densification and redundant primitives, degrading both efficiency and quality. Intuitively, this issue can be mitigated with scheduling strategies, which can be categorized into two paradigms: modulating target signal frequency via densification and modulating sampling frequency via image resolution. However, previous scheduling strategies are primarily hardcoded, failing to perceive the convergence behavior of scene frequency. To address this, we reframe the scene reconstruction problem from the perspective of signal structure recovery and propose SIG, a novel scheduler that synchronizes image supervision with Gaussian frequencies. Specifically, we derive the average sampling frequency and bandwidth of 3D representations, and then regulate the training image resolution and the Gaussian densification process based on scene frequency convergence. Furthermore, we introduce Sphere-Constrained Gaussians, which leverage the spatial prior of initialized point clouds to control Gaussian optimization. Our framework enables frequency-consistent, geometry-aware, and floater-free training, achieving state-of-the-art performance by a substantial margin in both efficiency and rendering quality in large-scale scenes. The code is available at: https://github.com/weiyixue999/Signal_Structure_Aware_Gaussian

Problem

• large-scale scene reconstruction
• sparse observations
• Gaussian Splatting
• uncontrolled densification
• redundant primitives

Innovation

• Structure-Aware Gaussian Splatting
• Signal Structure Recovery
• Frequency-Aware Scheduling
• Sphere-Constrained Gaussians
• Large-Scale Scene Reconstruction