AtomGS: Atomizing Gaussian Splatting for High-Fidelity Radiance Field

Rong Liu     Rui Xu     Yue Hu     Meida Chen     Andrew Feng

Abstract

3D Gaussian Splatting (3DGS) has recently advanced radiance field reconstruction by offering superior capabilities for novel view synthesis and real-time rendering speed. However, its strategy of blending optimization and adaptive density control might lead to sub-optimal results; it can sometimes yield noisy geometry and blurry artifacts due to prioritizing optimizing large Gaussians at the cost of adequately densifying smaller ones. To address this, we introduce AtomGS, consisting of Atomized Proliferation and Geometry-Guided Optimization. The Atomized Proliferation constrains ellipsoid Gaussians of various sizes into more uniform-sized Atom Gaussians. The strategy enhances the representation of areas with fine features by placing greater emphasis on densification in accordance with scene details. In addition, we proposed a Geometry-Guided Optimization approach that incorporates an Edge-Aware Normal Loss. This optimization method effectively smooths flat surfaces while preserving intricate details. Our evaluation shows that AtomGS outperforms existing state-of-the-art methods in rendering quality. Additionally, it achieves competitive accuracy in geometry reconstruction and offers a significant improvement in training speed over other SDF-based methods.

Atomized Proliferation

When handling input SfM points, 3DGS alternates between densification and optimization to enhance scene representation. In contrast, our method initially constrains Gaussians that represent fine details into Atom Gaussians and prioritizes their proliferation to quickly align with the scene's inherent geometry.

Drag the slider to play with proliferation! Also drag the comparison bar to see differences.

Iteration: 10

3DGS AtomGS (Ours)

Edge-Aware Normal Loss

We first compute the normal map using cross products of depth map gradients. Next, we derive the curvature map to show changes in surface normals, where higher values indicate rough textures and lower values mean smoother surfaces. However, optimizing this map can overly smooth details like sharp edges. To prevent this, we create an edge map from the RGB image's gradient magnitude. This helps preserve detailed features by preventing excessive smoothing in the curvature map, thus maintaining important geometric details. The illustration normal map is rendered from the 3DGS 30k result. Based on that, we can compute the normal loss map, showing the areas that our proposed loss could optimize.

Radiance Field Comparison

3DGS AtomGS (Ours) SuGaR

Mesh Comparison

SuGaR AtomGS (Ours) NeuS

BibTeX


  @misc{liu2024atomgs,
      title={AtomGS: Atomizing Gaussian Splatting for High-Fidelity Radiance Field}, 
      author={Rong Liu and Rui Xu and Yue Hu and Meida Chen and Andrew Feng},
      year={2024},
      eprint={2405.12369},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://rongliu-leo.github.io/AtomGS/}
  }