mlsanityData to 7 Oct 2026

StableGS: A Floater-Free Framework for 3D Gaussian Splatting

arXiv 2503.18458 ↗First posted 24 Mar 2025cs.CV cs.CLcited by 6

Not on the map. Why?

Abstract (arXiv)

3D Gaussian Splatting (3DGS) reconstructions are plagued by stubborn ``floater" artifacts that degrade their geometric and visual fidelity. We are the first to reveal the root cause: a fundamental conflict in the 3DGS optimization process where the opacity gradients of floaters vanish when their blended color reaches a pseudo-equilibrium of canceling errors against the background, trapping them in a spurious local minimum. To resolve this, we propose StableGS, a novel framework that decouples geometric regularization from final appearance rendering. Its core is a Dual Opacity architecture that creates two separate rendering paths: a ``Geometric Regularization Path" to bear strong depth-based constraints for structural correctness, and an ``Appearance Refinement Path" to generate high-fidelity details upon this stable foundation. We complement this with a synergistic set of geometric constraints: a self-supervised depth consistency loss and an external geometric prior enabled by our efficient global scale optimization algorithm. Experiments on multiple benchmarks show StableGS not only eliminates floaters but also resolves the common blur-artifact trade-off, achieving state-of-the-art geometric accuracy and visual quality.

Key facts. Cited by 6 papers in the map (data through 7 Oct 2026). Builds on 27 of its 31 references that are in the map. First posted on arXiv 24 Mar 2025.

Builds on

27 of 31 references.

References · 27

  1. 2018The Unreasonable Effectiveness of Deep Features as a Perceptual Metriccited by 11,029
  2. 20233D Gaussian Splatting for Real-Time Radiance Field Renderingcited by 6,141
  3. 2022Instant Neural Graphics Primitives with a Multiresolution Hash Encodingcited by 3,698
  4. 2021Mip-NeRF 360: Unbounded Anti-Aliased Neural Radiance Fieldscited by 2,013
  5. 2021Mip-NeRF: A Multiscale Representation for Anti-Aliasing Neural Radiance Fieldscited by 1,957
  6. 2023DUSt3R: Geometric 3D Vision Made Easycited by 1,526
  7. 2022TensoRF: Tensorial Radiance Fieldscited by 1,275
  8. 20242D Gaussian Splatting for Geometrically Accurate Radiance Fieldscited by 1,061
  9. 2021Direct Voxel Grid Optimization: Super-fast Convergence for Radiance Fields Reconstructioncited by 961
  10. 2024Grounding Image Matching in 3D with MASt3Rcited by 960
  11. 2020NeRF++: Analyzing and Improving Neural Radiance Fieldscited by 857
  12. 2023Zip-NeRF: Anti-Aliased Grid-Based Neural Radiance Fieldscited by 667
15 more
  1. 2023Mip-Splatting: Alias-free 3D Gaussian Splattingcited by 661
  2. 2023SuGaR: Surface-Aligned Gaussian Splatting for Efficient 3D Mesh Reconstruction and High-Quality Mesh Renderingcited by 611
  3. 20163DMatch: Learning Local Geometric Descriptors from RGB-D Reconstructionscited by 333
  4. 2024PGSR: Planar-based Gaussian Splatting for Efficient and High-Fidelity Surface Reconstructioncited by 280
  5. 2024DNGaussian: Optimizing Sparse-View 3D Gaussian Radiance Fields with Global-Local Depth Normalizationcited by 256
  6. 2024Gaussian Opacity Fields: Efficient Adaptive Surface Reconstruction in Unbounded Scenescited by 226
  7. 2024Global Structure-from-Motion Revisitedcited by 215
  8. 2024A Hierarchical 3D Gaussian Representation for Real-Time Rendering of Very Large Datasetscited by 214
  9. 2024Taming 3DGS: High-Quality Radiance Fields with Limited Resourcescited by 198
  10. 2023Depth-Regularized Optimization for 3D Gaussian Splatting in Few-Shot Imagescited by 172
  11. 2024Splatt3R: Zero-shot Gaussian Splatting from Uncalibrated Image Pairscited by 171
  12. 2024RaDe-GS: Rasterizing Depth in Gaussian Splattingcited by 110
  13. 2024GSDF: 3DGS Meets SDF for Improved Rendering and Reconstructioncited by 81
  14. 2024AbsGS: Recovering Fine Details for 3D Gaussian Splattingcited by 80
  15. 2024Mode-GS: Monocular Depth Guided Anchored 3D Gaussian Splatting for Robust Ground-View Scene Renderingcited by 5

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 4; 2026: 2.

6 papers.