We introduce Gaussian-enhanced Surfels (GESs), a bi-scale representation for radiance field rendering, wherein a set of 2D opaque surfels with view-dependent colors represent the coarse-scale geometry and appearance of scenes, and a few 3D Gaussians surrounding the surfels supplement fine-scale appearance details. The rendering with GESs consists of two passes -- surfels are first rasterized through a standard graphics pipeline to produce depth and color maps, and then Gaussians are splatted with depth testing and color accumulation on each pixel order independently. The optimization of GESs from multi-view images is performed through an elaborate coarse-to-fine procedure, faithfully capturing rich scene appearance. The entirely sorting-free rendering of GESs not only achieves very fast rates, but also produces view-consistent images, successfully avoiding popping artifacts under view changes. The basic GES representation can be easily extended to achieve anti-aliasing in rendering (Mip-GES), boosted rendering speeds (Speedy-GES) and compact storage (Compact-GES), and reconstruct better scene geometries by replacing 3D Gaussians with 2D Gaussians (2D-GES). Experimental results show that GESs advance the state-of-the-arts as a compelling representation for ultra-fast high-fidelity radiance field rendering.
The rendering of GES radiance fields consists of two passes, and is entirely sorting-free. Firstly, the opaque surfels are rasterized through a standard graphics pipeline, producing the color and depth maps. Secondly, we splat the Gaussians to the screen with depth testing, and accumulate the Gaussian color weighted by opacity on each pixel of the surfel color map, in an order-independent way. For each pixel, the Gaussians whose center depths fail to pass the depth testing with the surfel depth map will not accumulate color on the pixel, which means Gaussians are occluded by the geometry represented by the surfels. Such a sorting-free rendering not only bypasses the computation bottleneck of Gaussian sorting and achieves very fast frame rates, but also successfully avoids popping artifacts under view changes
The basic GES representation can be easily extended to further enhance its capability by incorporating recent improvements of the vanilla 3DGS method. By combining the filtering algorithm for Gaussian splatting [Yu et al. 2024] with the standard multi sampling anti-aliasing (MSAA) for surfel rasterization, we are able to significantly reduce aliasing artifacts in renderred images (Mip-GES). The Hessian pruning score [Hanson et al. 2024] can be employed to largely reduce the number of Gaussians, further boosting our rendering speed (Speedy-GES). Following [Lee et al. 2024], we can replace the SH coefficients of both surfels and Gaussians by querying colors from a hash grid, and quantize the scaling and rotation of surfels and Gaussians, obtaining a compact storage of GES (Compact-GES). We can also replace the 3D Gaussians with 2D Gaussians [Huang et al. 2024] to reconstruct better scene geometries (2D-GES).
*The quantitative results of rendering quality, speed and storage are evaluated on Mip-NeRF360 dataset [Barron et al. 2022] and the quantitative results of mesh reconstruction are evaluated on DTU dataset [Jensen et al. 2014].
[Kerbl et al. 2023] Bernhard Kerbl, Georgios Kopanas, Thomas Leimkühler, and George Drettakis. 2023. 3D Gaussian Splatting for Real-Time Radiance Field Rendering. ACM Transactions on Graphics (SIGGRAPH Conference Proceedings) 42, 4 (July 2023).
[Hanson et al. 2024] Alex Hanson, Allen Tu, Geng Lin, Vasu Singla, Matthias Zwicker, and Tom Goldstein. 2024. Speedy-Splat: Fast 3D Gaussian Splatting with Sparse Pixels and Sparse Primitives. arXiv preprint arXiv:2412.00578 (2024).
[Radl et al. 2024] Lukas Radl, Michael Steiner, Mathias Parger, Alexander Weinrauch, Bernhard Kerbl, and Markus Steinberger. 2024. StopThePop: Sorted Gaussian Splatting for View Consistent Real-time Rendering. ACM Trans. Graph. 43, 4, Article 64 (July 2024), 17 pages.
[Yu et al. 2024] Zehao Yu, Anpei Chen, Binbin Huang, Torsten Sattler, and Andreas Geiger. 2024a. Mip-Splatting: Alias-Free 3D Gaussian Splatting. In 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). 19447-19456.
[Lee et al. 2024] Joo Chan Lee, Daniel Rho, Xiangyu Sun, Jong Hwan Ko, and Eunbyung Park. 2024. Compact3dgaussianrepresentationforradiance field. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 21719-21728.
[Huang et al. 2024] Binbin Huang, Zehao Yu, Anpei Chen, Andreas Geiger, and Shenghua Gao. 2024. 2D Gaussian Splatting for Geometrically Accurate Radiance Fields. In ACM SIGGRAPH 2024 Conference Papers (Denver, CO, USA) (SIGGRAPH '24). Article 32, 11 pages.
[Barron et al. 2022] Jonathan T. Barron, Ben Mildenhall, Matthew Tancik, Peter Hedman, Ricardo Martin Brualla, and Pratul P. Srinivasan. 2021. Mip-NeRF: A Multiscale Representation for Anti-Aliasing Neural Radiance Fields. In 2021 IEEE/CVF International Conference on Computer Vision (ICCV). 5835-5844.
[Jensen et al. 2014] Rasmus Jensen, Anders Dahl, George Vogiatzis, Engin Tola, and Henrik Aanæs. 2014. Large scale multi-view stereopsis evaluation. In Proceedings of the IEEE conference on computer vision and pattern recognition. 406-413.
@article{ye2025gaussian,
title={When gaussian meets surfel: Ultra-fast high-fidelity radiance field rendering},
author={Ye, Keyang and Shao, Tianjia and Zhou, Kun},
journal={ACM Transactions on Graphics (TOG)},
volume={44},
number={4},
pages={1--15},
year={2025},
publisher={ACM New York, NY, USA}
}