AtomGS: Atomizing Gaussian Splatting for High-Fidelity Radiance Field

TL;DR

AtomGS is introduced, consisting of Atomized Proliferation and Geometry-Guided Optimization, which achieves competitive accuracy in geometry reconstruction and offers a significant improvement in training speed over other SDF-based methods.

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. More interactive demos can be found in our website (https://rongliu-leo.github.io/AtomGS/).

Figures (10)

• Figure 1: AtomGS outperforms existing methods in rendering quality and achieves competitive results in geometry accuracy by constraining Gaussians into Atom Gaussians and aligning them precisely with the natural geometry.
• Figure 2: Rendering Comparison of 7k Results: Ours vs. 3DGS. We display images using both full-size and shrunken Gaussians, examining the rendering effects and Gaussian placements. Our approach results in more precise geometric alignments, visible in fine details like bicycle spokes and blades of grass.
• Figure 3: Illustration of the Edge-Aware Normal Loss. The image maps are rendered from the 3DGS 30k result without applying our proposed Edge-Aware Normal Loss. Loss map (d) highlights the areas where it can improve the geometry over the 3DGS result.
• Figure 4: Radiance Field Comparison on the Mip-NeRF360 Dataset.
• Figure 5: Mesh Comparison on the DTU and NeRF Synthetic Datasets aanaes2016largemildenhall2021nerf