Research map / 3D vision / Depth and stereo
Camera Pose Estimation Using Implicit Distortion Models
Linfei Pan, Marc Pollefeys, Viktor Larsson
Check 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 |
|---|---|---|---|
| Jointly optimize poses across multiple images using the implicit distortion model | Camera pose estimation · Structure-from-Motion: Kirchenge, Grossmunster, Kazan, Doge Palace | position error: Kirchenge: 0.6 -> 0.4; Grossmunster: 0.9 -> 0.5; Kazan: 8.7 -> 3.4; Doge Palace: 3.4 -> 3.4 | Improved |
| Jointly optimize multiple images rather than estimate each image independently | Implicit distortion model · Aachen Day-Night and InLoc | localized images (%): Aachen day: 58.3 / 76.5 / 94.2 -> 82.6 / 92.4 / 98.3; night: 61.2 / 77.6 / 99.0 -> 73.5 / 88.8 / 100.0; InLoc duc1: 28.3 | Improved |
| Apply the proposed implicit calibration and pose estimation | Visual localization · Aachen Day-Night and InLoc | localized images (%): Aachen day: 82.6 / 92.4 / 98.3 vs 89.6 / 95.4 / 98.8; InLoc duc1: 34.8 / 52.5 / 69.7 vs 46.5 / 66.2 / 78.3 | Mixed |
| Filter spurious inliers (ablation) | Proposed implicit distortion model · Aachen Day-Night | localized images (%): day: 77.8 / 90.8 / 98.3 -> 82.6 / 92.4 / 98.3; night: 73.5 / 88.8 / 100.0 -> 73.5 / 88.8 / 100.0 | Improved |
| Filter spurious inliers (ablation) | Proposed implicit distortion model · InLoc | localized images (%): duc1: 35.4 / 53.0 / 69.7 -> 34.8 / 52.5 / 69.7; duc2: 35.9 / 58.0 / 74.0 -> 38.9 / 57.3 / 74.0 | Mixed |
| Use adjacent-point differences instead of local linear fitting (ablation) | Implicit distortion regularizer · Checkerboard calibration test sets | mean RMS reprojection error (px): 0.847 -> 6.881 | Hurt |
| Fit a quadratic rather than a linear local function (ablation) | Implicit distortion regularizer · Checkerboard calibration test sets | mean RMS reprojection error (px): 0.847 -> 1.208 | Hurt |
| Use 2 rather than 4 neighboring points (ablation) | Implicit distortion regularizer · Checkerboard calibration test sets | mean RMS reprojection error (px): 0.847 -> 0.881 | Hurt |
| Use 6 rather than 4 neighboring points (ablation) | Implicit distortion regularizer · Checkerboard calibration test sets | mean RMS reprojection error (px): 0.847 -> 0.839 | Improved |
| Use 8 rather than 4 neighboring points (ablation) | Implicit distortion regularizer · Checkerboard calibration test sets | mean RMS reprojection error (px): 0.847 -> 0.841 | Improved |
| Use 10 rather than 4 neighboring points (ablation) | Implicit distortion regularizer · Checkerboard calibration test sets | mean RMS reprojection error (px): 0.847 -> 0.842 | Improved |
| Use ℓ2 rather than Huber loss (ablation) | Implicit distortion regularizer · Checkerboard calibration test sets | mean RMS reprojection error (px): 0.847 -> 0.853 | Hurt |
Closest papers by meaning
- PRaDA: Projective Radial Distortion Averaging (CVPR 2025 · similarity 0.87)
- Parameter-Free Lens Distortion Calibration of Central Cameras (ICCV 2017 · similarity 0.87)
- AnyMap: Learning a General Camera Model for Structure-from-Motion with Unknown Distortion in Dynamic Scenes (CVPR 2025 · similarity 0.86)
- Four-View Geometry With Unknown Radial Distortion (CVPR 2023 · similarity 0.86)
- From Correspondences to Pose: Non-minimal Certifiably Optimal Relative Pose without Disambiguation (CVPR 2024 · similarity 0.84)
- Relative Pose Estimation through Affine Corrections of Monocular Depth Priors (CVPR 2025 · similarity 0.84)
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.
