Sparse-View Surface Reconstruction using Gaussian Splatting through High-Confidence Depth Propagation with Normal Priors

2026-07-04

AI Summary

This work addresses the challenges of 3D surface reconstruction from sparse viewpoints, where insufficient geometric cues and the discrete nature of Gaussian splats lead to depth discontinuities. Building upon 3D Gaussian Splatting (3DGS), the authors propose a normal-prior-guided, high-confidence depth propagation mechanism that diffuses reliable depth estimates into low-confidence regions. Additionally, they introduce an outlier-aware edge-preserving regularization term to mitigate depth artifacts caused by splat discreteness. The proposed method significantly enhances geometric completeness and accuracy of reconstructed surfaces under sparse-view settings, outperforming state-of-the-art approaches on the DTU and Tanks-and-Temples benchmarks while achieving high-fidelity surface reconstruction.

Abstract

3D reconstruction from sparse views is a challenging task in 3D computer vision. Recent studies on 3D Gaussian Splatting (3DGS) have achieved remarkable results with sparse views in novel view synthesis, yet reconstructing high-quality geometric surfaces from sparse views remains a challenge, due to the limited geometry clues and the discreteness of Gaussians. In this paper, we propose a novel 3DGS-based method for high-fidelity surface reconstruction from sparse views. Our key insight is to introduce a normal-guided depth propagation approach, which can extend depth information from high-confidence regions to constrain the depth in low-confidence areas. Additionally, we propose an abnormal depth edge-aware regularization to address depth discontinuities caused by the discreteness of Gaussians. Extensive experiments on DTU and Tanks-and-Temples datasets demonstrate that our method outperforms the state-of-the-art methods in sparse view surface reconstruction. Project page: https://hanl2010.github.io/DP-GS.

Problem

• Sparse-View Reconstruction
• Surface Reconstruction
• 3D Gaussian Splatting
• Depth Propagation
• Geometric Discreteness

Innovation

• Gaussian Splatting
• Sparse-View Reconstruction
• Depth Propagation
• Normal Priors
• Surface Reconstruction