[CVPR 2025] Luminance-GS: Adapting 3D Gaussian Splatting to Challenging Lighting Conditions with View-Adaptive Curve Adjustment

(Paper Link), (Extended Journal Link), (Website Link).

2026.5.10 : Thanks for the open issue, I fixed some bugs in Luminance-GS training.

2026.2.25 : The extended version, Luminance-GS++, has been released, link here!

2025.2.27 : Paper accepted by CVPR 2025 !

2025.4.27 : Code and dataset released !

Luminance-GS: Adapting 3D Gaussian Splatting to Challenging Lighting Conditions with View-Adaptive Curve Adjustment (Conference Version)

Ziteng Cui*¹, Xuangeng Chu¹, Tatsuya Harada^1,2.

^1.The University of Tokyo, ^2.RIKEN AIP.

Unifying Color and Lightness Correction with View-Adaptive Curve Adjustment for Robust 3D Novel View Synthesis (Journal Version)

Ziteng Cui¹, Shuhong Liu¹ Xiaoyu Dong¹ Xuangeng Chu¹, Lin Gu*² Ming-Hsuan Yang³ Tatsuya Harada^1,4.

^1.The University of Tokyo, ^2.Tohoku University, ^3.UC Merced, ^4.RIKEN AIP.

Use The Code:

cd Luminance-GS

Citation:

 @inproceedings{cui_luminance_gs,
	  title = {Luminance-GS: Adapting 3D Gaussian Splatting to Challenging Lighting Conditions with View-Adaptive Curve Adjustment},
	  author = {Cui, Ziteng and Chu, Xuangeng and Harada, Tatsuya},
	  booktitle={CVPR},
	  year={2025}}

@misc{cui_luminance_gs_plus_plus,
      title={Unifying Color and Lightness Correction with View-Adaptive Curve Adjustment for Robust 3D Novel View Synthesis}, 
      author={Ziteng Cui and Shuhong Liu and Xiaoyu Dong and Xuangeng Chu and Lin Gu and Ming-Hsuan Yang and Tatsuya Harada},
      year={2026},
      eprint={2602.18322},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2602.18322}, 
}

Dataset Citation:

@inproceedings{cui_aleth_nerf,
	  title={Aleth-NeRF: Illumination Adaptive NeRF with Concealing Field Assumption},
	  author={Cui, Ziteng and Gu, Lin and Sun, Xiao and Ma, Xianzheng and Qiao, Yu and Harada, Tatsuya},
	  booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
	  year={2024}
}

📖: Some Also Great Co-current Works (almost 3DGS in dark):

sRGB-based:

[PG 2024] Gaussian in the Dark: Real-Time View Synthesis From Inconsistent Dark Images Using Gaussian Splatting

[IROS 2024] DarkGS: Learning Neural Illumination and 3D Gaussians Relighting for Robotic Exploration

[CVPR 2025] lita-gs: illumination-agnostic novel view synthesis via reference-free 3d gaussian splatting and physical priors

[Arxiv 2025] LL-Gaussian: Low-Light Scene Reconstruction and Enhancement via Gaussian Splatting for Novel View Synthesis

RAW-based:

[NIPS 2024] Lighting Every Darkness with 3DGS: Fast Training and Real-Time Rendering for HDR View Synthesis

[NIPS 2024] From Chaos to Clarity: 3DGS in the Dark