Research map / 3D vision / Depth and stereo
Into the Twilight Zone: Depth Estimation using Joint Structure-Stereo Optimization
Aashish Sharma, Loong-Fah Cheong
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| replace DnCNN denoising + MeshStereo with joint structure-stereo optimization | DnCNN+MS · Middlebury v3, noise 60, 1px | bad pixel percentage: 58.97 -> 56.59 | Improved |
| replace BM3D denoising + MeshStereo with joint structure-stereo optimization | BM3D+MS · Middlebury v3, noise 60, 1px | bad pixel percentage: 63.35 -> 56.59 | Improved |
| replace SS-PCA joint denoising with joint structure-stereo optimization | SS-PCA · Middlebury v3, noise 60, 1px | bad pixel percentage: 69.17 -> 56.59 | Improved |
| use joint structure-stereo optimization at low noise instead of denoising baseline | DnCNN+MS · Middlebury v3, noise 25, 1px | bad pixel percentage: 43.43 -> 47.74 | Hurt |
| use joint structure-stereo optimization at low noise instead of denoising baseline | DnCNN+MS · Middlebury v3, noise 25, 3px | bad pixel percentage: 22.04 -> 25.12 | Hurt |
| use joint structure-stereo optimization at medium noise | DnCNN+MS · Middlebury v3, noise 50, 1px | bad pixel percentage: 55.06 -> 54.19 | Improved |
| use joint structure-stereo optimization at noise 60, 3px | DnCNN+MS · Middlebury v3, noise 60, 3px | bad pixel percentage: 32.68 -> 30.48 | Improved |
| use joint structure-stereo optimization on poorly-lit real night images | DnCNN+MS · Oxford RobotCar Set1, 5px | bad pixel percentage: 19.66 -> 12.31 | Improved |
| use joint structure-stereo optimization on poorly-lit real night images | DnCNN+MS · Oxford RobotCar Set1, 1px | bad pixel percentage: 63.86 -> 58.76 | Improved |
| use joint structure-stereo optimization on well-lit real night images | DnCNN+MS · Oxford RobotCar Set2, 2px | bad pixel percentage: 28.82 -> 28.80 | No clear effect |
| fine tune parameters on Set2 | Joint Structure-Stereo · Oxford RobotCar Set2, 1px | bad pixel percentage: 57.76 -> 56.45 | Improved |
| fine tune parameters on Set2 | Joint Structure-Stereo · Oxford RobotCar Set2, 5px | bad pixel percentage: 6.82 -> 6.08 | Improved |
Closest papers by meaning
- HyperDepth: Learning Depth From Structured Light Without Matching (CVPR 2016 · similarity 0.88)
- Single View Stereo Matching (CVPR 2018 · similarity 0.87)
- StereoGS: Sparse-View 3D Gaussian Splatting via Stereo Priors (ECCV 2026 · similarity 0.87)
- LoS: Local Structure-Guided Stereo Matching (CVPR 2024 · similarity 0.87)
- $\alpha$Depth: Learning Single-Pass Soft Boundary Decomposition for Stereo Conversion (NeurIPS 2026 · similarity 0.87)
- MonSter: Marry Monodepth to Stereo Unleashes Power (CVPR 2025 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
