SmokeSeer: 3D Gaussian Splatting for Smoke Removal and Scene Reconstruction

Neham Jain, Andrew Jong, Sebastian Scherer, Ioannis Gkioulekas

arXiv 2025

SmokeSeer teaser
Our method uses RGB and thermal images to jointly reconstruct the 3D scene and remove smoke using 3D Gaussian splatting.

Abstract

The presence of smoke in real-world scenes can severely degrade the quality of images and hamper visibility. Recently introduced methods for image restoration either rely on data-driven priors that are susceptible to hallucinations, or are limited to static low-density smoke. We introduce SmokeSeer, a method for performing simultaneous 3D scene reconstruction and smoke removal from a video capturing multiple views of a scene. To achieve this task, our method uses thermal and RGB images, leveraging the fact that the reduced scattering in thermal images enables us to see through the smoke. We build upon 3D Gaussian splatting to fuse information from the two image modalities, and decompose the scene explicitly into smoke and non-smoke components. Unlike prior approaches, SmokeSeer handles a broad range of smoke densities and can adapt to temporally varying smoke. We validate our approach on synthetic data and introduce a new real-world multi-view smoke dataset with RGB and thermal images.

Method

Method overview
An overview of SmokeSeer: (1) camera pose estimation and smoke segmentation, (2) initial surface reconstruction from thermal images, and (3) joint optimization of surface and smoke using RGB and thermal images.
Scattering vs wavelength
Scattering coefficient is significantly higher in the visible spectrum than in the long-wave infrared, motivating the use of thermal imaging for seeing through smoke.

Results

Thermal RGB Results Ground Truth
Thermal RGB Results Ground Truth
Thermal RGB Results Ground Truth
Thermal RGB Results Ground Truth

Resources

 Paper: Our paper is available on arXiv and locally.

 Code: Our code is available on Github.

 Data: Our data is available on Google Drive.

Citation

@article{jain2025smokeseer,
	title={{SmokeSeer: 3D Gaussian Splatting for Smoke Removal and Scene Reconstruction}},
	author={Neham Jain and Andrew Jong and Sebastian Scherer and Ioannis Gkioulekas},
	year={2025},
	eprint={2509.17329},
	archivePrefix={arXiv},
	primaryClass={cs.CV},
	url={https://arxiv.org/abs/2509.17329},
}

Acknowledgments

This work was supported by National Institute of Food and Agriculture award 2023-67021-39073, and Alfred P. Sloan Research Fellowship FG202013153 for Ioannis Gkioulekas.