Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

LSD3K: A Benchmark for Smoke Removal from Laparoscopic Surgery Images

Wenhui Chang, Hongming Chen

TL;DR

This work addresses the challenge of smoke obscuring laparoscopic views by introducing LSD3K, a public benchmark of 3,000 synthetic smoky image pairs generated with Blender-based non-uniform smoke over diverse Cholec80 backgrounds. The authors detail a reproducible synthesis pipeline, including density levels and ground-truth backgrounds, and validate the dataset with a user study showing realism close to real smoke. They benchmark representative desmoking methods, finding GridDehazeNet to offer the best quantitative performance and DehazeFormer to excel among transformer-based approaches, while highlighting the practical impact of smoke on downstream tasks like instrument tracking. Overall, LSD3K provides a scalable, high-quality dataset to accelerate the development of robust desmoking algorithms for surgical visualization and planning.

Abstract

Smoke generated by surgical instruments during laparoscopic surgery can obscure the visual field, impairing surgeons' ability to perform operations accurately and safely. Thus, smoke removal task for laparoscopic images is highly desirable. Despite laparoscopic image desmoking has attracted the attention of researchers in recent years and several algorithms have emerged, the lack of publicly available high-quality benchmark datasets is the main bottleneck to hamper the development progress of this task. To advance this field, we construct a new high-quality dataset for Laparoscopic Surgery image Desmoking, named LSD3K, consisting of 3,000 paired synthetic non-homogeneous smoke images. In this paper, we provide a dataset generation pipeline, which includes modeling smoke shape using Blender, collecting ground-truth images from the Cholec80 dataset, random sampling of smoke masks and etc. Based on the proposed benchmark, we further conducted a comprehensive evaluation of the existing representative desmoking algorithms. The proposed dataset is publicly available at https://drive.google.com/file/d/1v0U5_3S4nJpaUiP898Q0pc-MfEAtnbOq/view?usp=sharing

LSD3K: A Benchmark for Smoke Removal from Laparoscopic Surgery Images

TL;DR

This work addresses the challenge of smoke obscuring laparoscopic views by introducing LSD3K, a public benchmark of 3,000 synthetic smoky image pairs generated with Blender-based non-uniform smoke over diverse Cholec80 backgrounds. The authors detail a reproducible synthesis pipeline, including density levels and ground-truth backgrounds, and validate the dataset with a user study showing realism close to real smoke. They benchmark representative desmoking methods, finding GridDehazeNet to offer the best quantitative performance and DehazeFormer to excel among transformer-based approaches, while highlighting the practical impact of smoke on downstream tasks like instrument tracking. Overall, LSD3K provides a scalable, high-quality dataset to accelerate the development of robust desmoking algorithms for surgical visualization and planning.

Abstract

Smoke generated by surgical instruments during laparoscopic surgery can obscure the visual field, impairing surgeons' ability to perform operations accurately and safely. Thus, smoke removal task for laparoscopic images is highly desirable. Despite laparoscopic image desmoking has attracted the attention of researchers in recent years and several algorithms have emerged, the lack of publicly available high-quality benchmark datasets is the main bottleneck to hamper the development progress of this task. To advance this field, we construct a new high-quality dataset for Laparoscopic Surgery image Desmoking, named LSD3K, consisting of 3,000 paired synthetic non-homogeneous smoke images. In this paper, we provide a dataset generation pipeline, which includes modeling smoke shape using Blender, collecting ground-truth images from the Cholec80 dataset, random sampling of smoke masks and etc. Based on the proposed benchmark, we further conducted a comprehensive evaluation of the existing representative desmoking algorithms. The proposed dataset is publicly available at https://drive.google.com/file/d/1v0U5_3S4nJpaUiP898Q0pc-MfEAtnbOq/view?usp=sharing
Paper Structure (12 sections, 3 equations, 5 figures, 1 table)

This paper contains 12 sections, 3 equations, 5 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related Work
  3. Dataset Construction
  4. Smoke Synthesis
  5. Dataset Statistics
  6. User Study
  7. Application
  8. Algorithm Benchmarking
  9. Quantitative Evaluation
  10. Qualitative Evaluation
  11. Discussion
  12. Conclusion

Figures (5)

  • Figure 1: Illustration of the dataset generation pipeline.
  • Figure 2: More example image pairs sampled from the proposed LSD3K. Best viewed by zooming in the figures.
  • Figure 3: User study results. The ratings given by all participants on different smoke datasets.
  • Figure 4: Visual results of paired image instrument tracking.
  • Figure 5: Visual results of paired image instrument tracking.