Research map / Point clouds / Registration and matching
Efficient and Robust Registration on the 3D Special Euclidean Group
Uttaran Bhattacharya, Venu Madhav Govindu
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 |
|---|---|---|---|
| use 3 IRLS steps per iteration instead of 1 (ablation) | Our L1/2 · Pair of ICL-NUIM scans, convergence eps 1e-7 | outer iterations to converge: 46 -> 18 | Improved |
| use 2 IRLS steps per iteration instead of 1 (ablation) | Our L1/2 · Pair of ICL-NUIM scans, convergence eps 1e-7 | outer iterations to converge: 46 -> 25 | Improved |
| use intrinsic Lie algebra parametrization instead of extrinsic parametrization of the line process (ablation) | Our L1/2 (single IRLS step) · Pair of ICL-NUIM scans, convergence eps 1e-7 | outer iterations to converge: 63 -> 46 | Improved |
| use L1/2 loss instead of scaled Geman-McClure (ablation) | Our method · Augmented ICL-NUIM livingroom 1 (pairwise) | median rotation angle error (deg): 0.793 -> 0.595 | Improved |
| use L1/2 loss instead of L1 (ablation) | Our method · Synthetic range data (sigma=0.005) | median rotation angle error (deg): 1.516 -> 0.959 | Improved |
| use L1 loss instead of L1/2 (ablation) | Our method · Augmented ICL-NUIM livingroom 1 (pairwise) | running time (ms) and median rotation error: 61 -> 33 ms; 0.595 -> 0.964 deg | Mixed |
| use L1/2 loss instead of L1 (ablation) | Our method · Augmented ICL-NUIM livingroom 1 (multiview) | mean registration error (m): 0.07 -> 0.04 | Improved |
| replace FGR line process with IRLS on SE(3) (L1/2) | FGR · Synthetic range data (mean over 5 models) | running time (ms): 17.4 -> 5.3 | Improved |
| replace FGR with our Geman-McClure IRLS | FGR · Augmented ICL-NUIM livingroom 1 (pairwise) | running time (ms): 272 -> 118 | Improved |
Closest papers by meaning
- Fast and Robust Non-Rigid Registration Using Accelerated Majorization-Minimization (TPAMI 2023 · similarity 0.89)
- Scalable 3D Registration via Truncated Entry-wise Absolute Residuals (CVPR 2024 · similarity 0.89)
- Group-Wise Point-Set Registration Based on Renyi's Second Order Entropy (CVPR 2017 · similarity 0.87)
- Fast and Robust Iterative Closest Point (TPAMI 2021 · similarity 0.87)
- Go-ICP: A Globally Optimal Solution to 3D ICP Point-Set Registration (TPAMI 2016 · similarity 0.87)
- Progressive Correspondence Regenerator for Robust 3D Registration (CVPR 2025 · similarity 0.86)
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.
