Towards Rotation-Only Imaging Geometry: Multi-View Pose Estimation
Organizations: Shanghai Key Laboratory of Navigation and Location-based Services, School of Automation and Intelligent Sensing, Shanghai Jiao Tong University, Shanghai 200240, China
Abstract
Global structure-from-motion (SfM) commonly refines relative poses on individual image pairs and then estimates global rotations by rotation averaging. Pairwise refinement does not explicitly combine feature-level constraints with multi-view consistency during rotation initialization. We introduce Global Rotations from Aggregated Matches (GRAM), which estimates global rotations directly from initial relative rotations and inlier matches by minimizing edge-wise rotation-only algebraic residuals. GRAM replaces two-view relative-pose optimization and rotation averaging with a single stage. We further extend rotation- only imaging geometry to multiple views and develop Multi-view Optimization with Rotation-only Reprojection Error (MORE) for nonlinear global rotation refinement. Analytical Rotation-guided Translation (ART) derives locally scaled triplet translations from Linear Global Translation (LiGT) constraints and recovers global translations using weighted least squares. Synthetic and real-world experiments demonstrate the accuracy, robustness, and efficiency of the proposed methods.
Figures & tables
| Global Rotation Initialization (deg) | Fixed-Rotation Global Translation (scene units) | Complete Pipelines | |||||||
| Scene (Cams) | GLOMAP-Avg [ 26 ] | GRAM | GLOMAP-GP [ 26 ] | LiGT [ 2 ] | ART | Rotation Error (deg) | Global-Translation Error (scene units) | ||
| GLOMAP [ 26 ] | Our Pipeline | GLOMAP [ 26 ] | Our Pipeline | ||||||
| Herz-Jesus-P25 (25) | 0.0397 | 0.0327 | 0.0105 | 0.0445 | 0.0092 | 0.0284 | 0.0219 | 0.0055 | 0.0054 |
| Herz-Jesus-P8 (8) | 0.0207 | 0.0115 | 0.0055 | 0.0050 | 0.0049 | 0.0151 | 0.0110 | 0.0033 | 0.0025 |
| Castle-P19 (19) | 0.0920 | 0.0831 | 0.7017 | 5.5522 | 0.0532 | 0.0341 | 0.0725 | 0.0196 | 0.0469 |
| Castle-P30 (30) | 0.1022 | 0.1008 | 0.3553 | 3.9241 | 0.0402 | 0.0410 | 0.0343 | 0.0204 | 0.0188 |
| GLOMAP | Ours | ||
|---|---|---|---|
| Stage | Time (s) | Stage | Time (s) |
| Sampson two-view refinement | 8.330 | GRAM | 0.042 |
| Rotation Avg. | 0.002 | MORE | 0.621 |
| GLOMAP-GP | 7.526 | ART | 0.043 |
| Triangulation | 0.935 | ||
| BA | 2.850 | ||
Camera Pose Estimation
Camera pose estimation recovers a camera’s position and orientation from images, visual features, scene geometry, and sometimes sensor measurements.
See all 264 papers