Research map / Image generation / Image restoration
Rectification Reimagined: A Unified Mamba Model for Image Correction and Rectangling with Prompts
Linwei Qiu, Gongzhe Li, Xiaozhe Zhang, Qilin Sun, Fengying Xie
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 |
|---|---|---|---|
| add visual prompts (ablation) | UniRect · Portrait correction (T1) | ShapeACC: 96.324 -> 97.454 | Improved |
| add visual prompts (ablation) | UniRect · Stitched image rectangling (T3) | PSNR: 23.82 -> 25.10 | Improved |
| add visual prompts (ablation) | UniRect · Rotation correction (T4) | PSNR: 23.15 -> 23.16 | No clear effect |
| add RP-TPS (ablation) | UniRect · Rotation correction (T4) | PSNR: 21.56 -> 23.16 | Improved |
| add restoration module (ablation) | UniRect · Rotation correction (T4) | PSNR: 22.84 -> 23.16 | Improved |
| add restoration module (ablation) | UniRect · Rectified wide-angle rectangling (T2) | PSNR: 19.99 -> 19.90 | Mixed |
| use SMoEs instead of mixed learning for four-in-one (ablation) | UniRect · Rectified wide-angle rectangling (T2) | PSNR: 13.07 -> 19.90 | Improved |
| use SMoEs instead of sequential learning (4-1-3-2) (ablation) | UniRect · Rotation correction (T4) | PSNR: 22.93 -> 23.16 | Improved |
| replace CNN with Mamba blocks (ablation) | UniRect · Stitched image rectangling (Table 5 block 1) | PSNR: 24.12 -> 25.10 | Improved |
| replace Transformer with Mamba blocks (ablation) | UniRect · Table 5 block 2 | PSNR: 23.23 -> 23.16 | Mixed |
| use unified UniRect instead of task-specific methods | portrait correction · Portrait correction (T1) | ShapeACC: 97.491 -> 97.454 | No clear effect |
Closest papers by meaning
- Lifting the Structural Morphing for Wide-Angle Images Rectification: Unified Content and Boundary Modeling (ICCV 2025 · similarity 0.88)
- Multi-Level Curriculum for Training a Distortion-Aware Barrel Distortion Rectification Model (ICCV 2021 · similarity 0.88)
- RecRecNet: Rectangling Rectified Wide-Angle Images by Thin-Plate Spline Model and DoF-based Curriculum Learning (ICCV 2023 · similarity 0.88)
- ReCoNet: Recurrent Correction Network for Fast and Efficient Multi-Modality Image Fusion (ECCV 2022 · similarity 0.86)
- MobileMamba: Lightweight Multi-Receptive Visual Mamba Network (CVPR 2025 · similarity 0.84)
- Rectification-specific Supervision and Constrained Estimator for Online Stereo Rectification (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.
