Experiments
Although 3D Gaussian Splatting (3DGS) has achieved im-pressive performance in real-time rendering, its densification strategy often results in suboptimal reconstruction quality. In this work, we present a comprehensive improvement to the densification pipeline of 3DGS from three perspectives: when to densify, how to densify, and how to mitigate overfitting. Specifically, we propose an Edge-Aware Score to effectively select candidate Gaussians for splitting. We further introduce a Long-Axis Split strategy that reduces geometric distortions introduced by clone and split operations. To address over-fitting, we design a set of techniques, including Recovery-Aware Pruning, Multi-step Update, and Growth Control. Our method enhances rendering fidelity without introducing addi-tional training or inference overhead, achieving state-of-the-art performance with fewer Gaussians.
@misc{deng2025improvingdensification3dgaussian,
title={Improving Densification in 3D Gaussian Splatting for High-Fidelity Rendering},
author={Xiaobin Deng and Changyu Diao and Min Li and Ruohan Yu and Duanqing Xu},
year={2025},
eprint={2508.12313},
archivePrefix={arXiv},
primaryClass={cs.CV},
url={https://arxiv.org/abs/2508.12313},
}