Research map / Image generation / Editing and style transfer
Multi-History-Step SDE Inversion for Image Editing with Superior Regional Awareness
Haiyan Wei, Yunlong Wang, Huaibo Huang, Zhenan Sun, Kunbo Zhang
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 |
|---|---|---|---|
| MIEdit (SD3) vs FlowEdit baseline on background preservation | MIEdit · PIE-Bench | PSNR: 26.56 -> 29.44 | Improved |
| MIEdit with SD3 vs FlowCycle efficiency | MIEdit · PIE-Bench | NFE: 6633 -> 38 | Improved |
| MIEdit (SD3) vs best baselines on human-aligned score | MIEdit · EditEval++ | EditScore: 4.05 -> 4.30 | Improved |
| increase history terms from (1,1) to (2,3) predictor/corrector (ablation) | MIEdit PixArt-Sigma · PIE-Bench | CLIP similarity: 21.73 -> 22.26 | Improved |
| increase history terms from (2,3) to (3,3) (ablation) | MIEdit PixArt-Sigma · PIE-Bench | PSNR / CLIP: PSNR 28.01 -> 28.08; CLIP 22.26 -> 22.17 | Mixed |
| IASM instead of InfEdit attention mask (ablation) | MIEdit SD3.5 · PIE-Bench | LPIPS: 48.06 -> 42.36 | Improved |
| IASM instead of Follow-Your-Shape sampling-generated mask (ablation) | MIEdit SD3.5 · PIE-Bench | PSNR: 26.82 -> 29.35 | Improved |
| cosine-angle mask instead of magnitude-based mask (ablation) | MIEdit SD3.5 · PIE-Bench | LPIPS: 44.63 -> 42.36 | Improved |
| temporal mask scheduling instead of full-step masks (ablation) | MIEdit SD3.5 · PIE-Bench | LPIPS: 43.20 -> 42.36 | Improved |
| power-interpolated beta schedule instead of linear (ablation) | MIEdit · large edits | image quality | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- FlowAlign — FlowAlign: Trajectory-Regularized, Inversion-Free Flow-based Image Editing (ICLR 2026) · better on 3 of 3 shared results
- FlowEdit — FlowEdit: Inversion-Free Text-Based Editing Using Pre-Trained Flow Models (ICCV 2025) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Inversion-Free Image Editing with Language-Guided Diffusion Models (CVPR 2024 · similarity 0.91)
- Tuning-Free Inversion-Enhanced Control for Consistent Image Editing (AAAI 2024 · similarity 0.90)
- DCI: Dual-Conditional Inversion for Boosting Diffusion-Based Image Editing (NeurIPS 2025 · similarity 0.90)
- PnP Inversion: Boosting Diffusion-based Editing with 3 Lines of Code (ICLR 2024 · similarity 0.90)
- Semantic Image Inversion and Editing using Rectified Stochastic Differential Equations (ICLR 2025 · similarity 0.89)
- Object-Aware Inversion and Reassembly for Image Editing (ICLR 2024 · similarity 0.89)
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.
