HoGS: Unified Near and Far Object Reconstruction via Homogeneous Gaussian Splatting

Abstract

Novel view synthesis has demonstrated impressive progress recently, with 3D Gaussian splatting (3DGS) offering efficient training time and photorealistic real-time rendering. However, reliance on Cartesian coordinates limits 3DGS's performance on distant objects, which is important for reconstructing unbounded outdoor environments.

We found that, despite its ultimate simplicity, using homogeneous coordinates, a concept on the projective geometry, for the 3DGS pipeline remarkably improves the rendering accuracies of distant objects. We therefore propose Homogeneous Gaussian Splatting (HoGS) incorporating homogeneous coordinates into the 3DGS framework, providing a unified representation for enhancing near and distant objects.

HoGS effectively manages both expansive spatial positions and scales particularly in outdoor unbounded environments by adopting projective geometry principles. Experiments show that HoGS significantly enhances accuracy in reconstructing distant objects while maintaining high-quality rendering of nearby objects, along with fast training speed and real-time rendering capability.

Video

Method Overview

HoGS uses homogeneous coordinates in the 3D Gaussian Splatting (3DGS) framework to represent both positions and scales. The nearby and distant Gaussian positions are represented seamlessly by homogeneous coordinates defined in the projective geometry, and the Gaussian's distance and scales change proportionally with the homogeneous component in the scale. Our HoGS simplifies the optimization of positions and scales of distant objects, and enables the accurate rendering of unbounded scenes while preserving the 3DGS’s fast training time and real-time performance.

Results

We evaluated our method on a total of 17 real scenes from multiple datasets containing diverse environments. In particular, we test our method on the full dataset of Mip-NeRF360, including four indoor and five outdoor scenes, two selected scenes from the Tanks&Temples dataset, as well as three indoor and three outdoor scenes from DL3DV benchmark dataset. For the full evaluation, please check the paper and the supplemental.

BibTeX

@inproceedings{liuHoGS,
      author       = {Xinpeng Liu, Zeyi Huang, Fumio Okura, Yasuyuki Matsushita},
      title        = {HoGS: Unified Near and Far Object Reconstruction via Homogeneous Gaussian Splatting},
      booktitle    = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
      year         = {2025}
}