Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Monocular Depth Guided Occlusion-Aware Disparity Refinement via Semi-supervised Learning in Laparoscopic Images

Ziteng Liu, Dongdong He, Chenghong Zhang, Wenpeng Gao, Yili Fu

TL;DR

This work tackles occlusion and data scarcity in stereo laparoscopic disparity estimation by introducing DGORNet, a monocular-depth-guided occlusion-aware refinement framework. It integrates a Depth Predictor with an Occlusion Mask Prediction Module and a Disparity Refinement Module, enhanced by a Position Embedding to supply explicit spatial context and a semi-supervised Optical Flow Difference Loss to leverage temporal continuity across frames. On the SCARED dataset, DGORNet achieves state-of-the-art End-Point Error ($EPE$) and RMSE, with ablations confirming the benefits of Position Embedding and OFDLoss in improving spatial and temporal consistency. The proposed approach offers a practical, robust solution for disparity refinement in laparoscopic surgery, delivering accuracy gains in occluded and textureless regions and supporting real-time deployment variants.

Abstract

Occlusion and the scarcity of labeled surgical data are significant challenges in disparity estimation for stereo laparoscopic images. To address these issues, this study proposes a Depth Guided Occlusion-Aware Disparity Refinement Network (DGORNet), which refines disparity maps by leveraging monocular depth information unaffected by occlusion. A Position Embedding (PE) module is introduced to provide explicit spatial context, enhancing the network's ability to localize and refine features. Furthermore, we introduce an Optical Flow Difference Loss (OFDLoss) for unlabeled data, leveraging temporal continuity across video frames to improve robustness in dynamic surgical scenes. Experiments on the SCARED dataset demonstrate that DGORNet outperforms state-of-the-art methods in terms of End-Point Error (EPE) and Root Mean Squared Error (RMSE), particularly in occlusion and texture-less regions. Ablation studies confirm the contributions of the Position Embedding and Optical Flow Difference Loss, highlighting their roles in improving spatial and temporal consistency. These results underscore DGORNet's effectiveness in enhancing disparity estimation for laparoscopic surgery, offering a practical solution to challenges in disparity estimation and data limitations.

Monocular Depth Guided Occlusion-Aware Disparity Refinement via Semi-supervised Learning in Laparoscopic Images

TL;DR

This work tackles occlusion and data scarcity in stereo laparoscopic disparity estimation by introducing DGORNet, a monocular-depth-guided occlusion-aware refinement framework. It integrates a Depth Predictor with an Occlusion Mask Prediction Module and a Disparity Refinement Module, enhanced by a Position Embedding to supply explicit spatial context and a semi-supervised Optical Flow Difference Loss to leverage temporal continuity across frames. On the SCARED dataset, DGORNet achieves state-of-the-art End-Point Error () and RMSE, with ablations confirming the benefits of Position Embedding and OFDLoss in improving spatial and temporal consistency. The proposed approach offers a practical, robust solution for disparity refinement in laparoscopic surgery, delivering accuracy gains in occluded and textureless regions and supporting real-time deployment variants.

Abstract

Occlusion and the scarcity of labeled surgical data are significant challenges in disparity estimation for stereo laparoscopic images. To address these issues, this study proposes a Depth Guided Occlusion-Aware Disparity Refinement Network (DGORNet), which refines disparity maps by leveraging monocular depth information unaffected by occlusion. A Position Embedding (PE) module is introduced to provide explicit spatial context, enhancing the network's ability to localize and refine features. Furthermore, we introduce an Optical Flow Difference Loss (OFDLoss) for unlabeled data, leveraging temporal continuity across video frames to improve robustness in dynamic surgical scenes. Experiments on the SCARED dataset demonstrate that DGORNet outperforms state-of-the-art methods in terms of End-Point Error (EPE) and Root Mean Squared Error (RMSE), particularly in occlusion and texture-less regions. Ablation studies confirm the contributions of the Position Embedding and Optical Flow Difference Loss, highlighting their roles in improving spatial and temporal consistency. These results underscore DGORNet's effectiveness in enhancing disparity estimation for laparoscopic surgery, offering a practical solution to challenges in disparity estimation and data limitations.
Paper Structure (13 sections, 6 equations, 4 figures, 2 tables)

This paper contains 13 sections, 6 equations, 4 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Method
  3. Network Structure
  4. Position Embedding
  5. Optical Flow Difference Loss
  6. Experiments
  7. Datasets and Metrics
  8. Implementation Details
  9. Loss Function
  10. Results and Discussion
  11. Quantitative Comparisons
  12. Ablation Study
  13. Conclusion

Figures (4)

  • Figure 1: The overview of our network. OPM is the Occlusion Mask Prediction Module. DRM is the Disparity Refinement Module. PE is the Position Embedding.
  • Figure 2: Error distribution map of disparity prediction result of five state-of-the-art methods on SCARED dataset. Brighter color indicates high error value.
  • Figure 3: An example of the relationship between a point's movement and disparity.
  • Figure 4: Prediction result on the SEK.