Seeing Through Smoke: Surgical Desmoking for Improved Visual Perception

Jingpei Lu; Fengyi Jiang; Xiaorui Zhang; Lingbo Jin; Omid Mohareri

cs.CVMar 26, 2026

Seeing Through Smoke: Surgical Desmoking for Improved Visual Perception

Authors: Jingpei Lu, Fengyi Jiang, Xiaorui Zhang, Lingbo Jin, Omid Mohareri

Organizations: Intuitive Surgical, Sunnyvale, CA 94086, USA

Abstract

Minimally invasive and robot-assisted surgery relies heavily on endoscopic imaging, yet surgical smoke produced by electrocautery and vessel-sealing instruments can severely degrade visual perception and hinder vision-based functionalities. We present a transformer-based surgical desmoking model with a physics-inspired desmoking head that jointly predicts smoke-free image and corresponding smoke map. To address the scarcity of paired smoky-to-smoke-free training data, we develop a synthetic data generation pipeline that blends artificial smoke patterns with real endoscopic images, yielding over 80,000 paired samples for supervised training. We further curate, to our knowledge, the largest paired surgical smoke dataset to date, comprising 5,817 image pairs captured with the da Vinci robotic surgical system, enabling benchmarking on high-resolution endoscopic images. Extensive experiments on both a public benchmark and our dataset demonstrate state-of-the-art performance in image reconstruction compared to existing dehazing and desmoking approaches. We also assess the impact of desmoking on downstream stereo depth estimation and instrument segmentation, highlighting both the potential benefits and current limitations of digital smoke removal methods.

Figures & tables

Figure 1: Overview of the proposed surgical desmoking model.

Figure 2: (a) Synthetic smoke data generation pipeline. (b) Qualitative results for desmoking and smoke-map prediction on real smoky images.

Dataset	Smoke type	Size	Image type (resolution)	Availability
LSD3K [ 4 ]	Synthetic	3,000 pairs	Mono (480 $\times$ 480)	✗
LSVD [ 27 ]	Real	786 clips	Mono (1080 $\times$ 1920)	✗
De-Smoking [ 28 ]	Real	3,464 pairs	Mono (inconsistent)	✓
Ours	Real	5,817 pairs	Stereo (1024 $\times$ 1280)	✓

Table 1: Existing surgical smoke datasets.

Figure 3: Qualitative comparisons on our test set across representative scenes (top: liver; bottom: stomach)

Method	Cholecystectomy		Prostatectomy
Method	SSIM	PSNR	SSIM	PSNR
He et al. [ 8 ]	0.73 $\pm$ 0.09	18.44 $\pm$ 3.03	0.72 $\pm$ 0.12	19.39 $\pm$ 3.38
Song et al. [ 21 ]	0.83 $\pm$ 0.10	23.25 $\pm$ 4.51	0.73 $\pm$ 0.15	19.22 $\pm$ 5.93
Qin et al. [ 17 ]	0.84 $\pm$ 0.10	24.49 $\pm$ 5.88	0.72 $\pm$ 0.15	19.13 $\pm$ 6.68
Salazar et al. [ 19 ]	0.75 $\pm$ 0.12	20.06 $\pm$ 4.00	0.72 $\pm$ 0.13	19.94 $\pm$ 4.10
Liu et al. [ 14 ]	0.81 $\pm$ 0.09	22.21 $\pm$ 3.79	0.73 $\pm$ 0.13	20.08 $\pm$ 5.70

Table 2: Quantitative results on the De-Smoking dataset

Figure 4: Qualitative results on downstream stereo depth estimation (top) and instrument segmentation (bottom) after different desmoking methods.

Method	Reconstruction		Depth	Segmentation
Method	SSIM $\uparrow$	PSNR $\uparrow$	MAE (mm) $\downarrow$	$\%$ IoU $\uparrow$
None	–	–	47.22 $\pm$ 61.48	81.15 $\pm$ 17.01
He et al. [ 8 ]	0.53 $\pm$ 0.07	12.83 $\pm$ 1.53	51.53 $\pm$ 66.71	80.81 $\pm$ 17.03
Song et al. [ 21 ]	0.61 $\pm$ 0.09	14.45 $\pm$ 1.83	49.00 $\pm$ 66.45	80.90 $\pm$ 16.27
Qin et al. [ 17 ]	0.72 $\pm$ 0.09	19.57 $\pm$ 3.20	49.03 $\pm$ 63.84	81.15 $\pm$ 17.15
Salazar et al. [ 19 ]	0.54 $\pm$ 0.12	14.21 $\pm$ 2.05	53.73 $\pm$ 66.50	79.46 $\pm$ 16.85

Table 3: Quantitative results on our test set for different desmoking methods. We report reconstruction quality (SSIM/PSNR) and downstream performance after desmoking: stereo depth estimation (MAE) and instrument segmentation (IoU).

Explore similar work

SurgiATM: A Physics-Guided Plug-and-Play Model for Deep Learning-Based Smoke Removal in Laparoscopic Surgery
Nov 7, 2025Mingyu Sheng, Jianan Fan, Dongnan Liu +3Physics-Guided Image Restoration Image Restoration
School of Computer Science, The University of Sydney, Sydney, NSW, Australia · Institute of Medical Robotics, School of Biomedical Engineering, Shanghai Jiao Tong University, Shanghai, China · Department of Radiology, Brigham and Women’s Hospital, and Harvard Medical School, Boston, MA, United States
PhySe-RPO: Physics and Semantics Guided Relative Policy Optimization for Diffusion-Based Surgical Smoke Removal
Mar 24, 2026Zining Fang, Cheng Xue, Chunhui Liu +3Image Restoration Diffusion Models for Image Restoration
School of Computer Science and Engineering, Southeast University · Zhongda Hospital,Southeast University · School of Future Technology, South China University of Technology
GenSmoke-GS: A Multi-Stage Method for Novel View Synthesis from Smoke-Degraded Images Using a Generative Model
Apr 3, 2026Qida Cao, Xinyuan Hu, Changyue Shi +3Novel View Synthesis Multi-View 3D Reconstruction
School of Computer Science and Technology, Hangzhou Dianzi University · School of Computer Science and Technology, Harbin Institute of Technology, Shenzhen

Nov 7, 2025J/K move · Enter open · S save

SurgiATM: A Physics-Guided Plug-and-Play Model for Deep Learning-Based Smoke Removal in Laparoscopic Surgery

Mingyu Sheng, Jianan Fan, Dongnan Liu, Guoyan Zheng+2

School of Computer Science, The University of Sydney, Sydney, NSW, Australia · Institute of Medical Robotics, School of Biomedical Engineering, Shanghai Jiao Tong University, Shanghai, China · Department of Radiology, Brigham and Women’s Hospital, and Harvard Medical School, Boston, MA, United States

Physics-Guided Image Restoration Image Restoration

Open paper AskPDF arXiv

During laparoscopic surgery, smoke generated by tissue cauterization can significantly degrade the visual quality of endoscopic frames, increasing the risk of surgical errors and hindering both clinical decision-making and computer-assisted visual analysis. Consequently, removing surgical smoke is critical to ensuring patient safety and maintaining operative efficiency. In this study, we propose the Surgical Atmospheric Model (SurgiATM) for surgical smoke removal. SurgiATM statistically bridges a physics-based atmospheric model and data-driven deep learning models, combining the superior generalizability of the former with the high accuracy of the latter. Furthermore, SurgiATM is designed as a lightweight module that can be easily integrated into existing surgical desmoking architectures with minimal modification, aiming to enhance their accuracy and stability. The proposed method is derived via statistically optimizing a Mixture-of-Experts (MoE) model at the output end of arbitrary deep learning methods, with a Laplacian-like error distribution specifically leveraged to model surgical smoke. The output-stage MoE ensures minimal modification to the architecture of the original methods, while the Laplacian-like distribution characteristic of surgical smoke enables a lightweight reconstruction formulation with minimal parameters. Therefore, SurgiATM introduces only two hyperparameters and no additional trainable weights, preserving the original network architecture with minimal computational and modification overhead. We conduct extensive experiments on three public surgical datasets with ten desmoking methods, involving multiple network architectures and covering diverse procedures, including cholecystectomy, partial nephrectomy, and diaphragm dissection.

Abstract

Explore similar work

SurgiATM: A Physics-Guided Plug-and-Play Model for Deep Learning-Based Smoke Removal in Laparoscopic Surgery

PhySe-RPO: Physics and Semantics Guided Relative Policy Optimization for Diffusion-Based Surgical Smoke Removal

GenSmoke-GS: A Multi-Stage Method for Novel View Synthesis from Smoke-Degraded Images Using a Generative Model