Research map / Point clouds / Point cloud classification
Fast and Interpretable 2D Homography Decomposition: Similarity-Kernel-Similarity and Affine-Core-Affine Transformations
Shen Cai, Zhanhao Wu, Lingxi Guo, Jiachun Wang, Siyu Zhang, Junchi Yan, Shuhan Shen
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 SKS decomposition instead of NDLT-SVD | 4-point homography solver · 4-point homography | FLOPs: about 162x speedup (27400 -> 169) | Improved |
| use SKS instead of HO-SVD | 4-point homography solver · 4-point homography | FLOPs: about 11x (1800 -> 169) | Improved |
| use SKS instead of GPT-LU | 4-point homography solver · 4-point homography | FLOPs: about 12x (1950 -> 169) | Improved |
| use SKS instead of RHO-GE | 4-point homography solver · 4-point homography | FLOPs: about 1.3x (223 -> 169) | Improved |
| use ACA instead of SKS | 4-point homography solver · 4-point homography | FLOPs: 169 -> 97 | Improved |
| use ACA instead of NDLT-SVD, HO-SVD, GPT-LU, RHO-GE | 4-point homography solver · 4-point homography | FLOPs fraction: 0.35%, 5.4%, 5.0% and 43% of their FLOPs | Improved |
| use ACA(*) instead of RHO-GE(*) at O1 | 4-point homography · CPU runtime O1 | runtime (us): 0.0327 -> 0.0146 | Improved |
| use ACA(*) instead of NDLT-SVD(**) at O2 | 4-point homography · CPU runtime O2 | runtime (us): 12.5 -> 0.0145 | Improved |
| use SKS(*) instead of RHO-GE(*) at O2 | 4-point homography · CPU runtime O2 | runtime (us): 0.0287 -> 0.0252 | No clear effect |
| use single instead of double precision for SKS at Od (ablation) | SKS · CPU runtime Od | runtime (us): 0.0901 -> 0.0595 | Improved |
| specialize ACA to a rectangular source plane (TensorACA) | ACA · rectangle source | FLOPs: 97 -> 47 | Improved |
| drop pivoting as in RHO-GE | RHO-GE · 4-point homography robustness | numerical robustness: fails for improper point order | Hurt |
Closest papers by meaning
- Homography Decomposition Revisited (IJCV 2026 · similarity 0.88)
- Noisy One-point Homographies are Surprisingly Good (CVPR 2024 · similarity 0.87)
- Unsupervised Global and Local Homography Estimation With Motion Basis Learning (TPAMI 2022 · similarity 0.87)
- Adapting Dense Matching for Homography Estimation with Grid-based Acceleration (CVPR 2025 · similarity 0.85)
- Mask-Homo: Pseudo Plane Mask-Guided Unsupervised Multi-Homography Estimation (AAAI 2024 · similarity 0.84)
- MCNet: Rethinking the Core Ingredients for Accurate and Efficient Homography Estimation (CVPR 2024 · 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.
