Research map / Image generation / Editing and style transfer
Wavelet-Guided Semantic Signal Compensation for Inversion-Free Image Editing
Anqi Tang, Wenhao Sun, Zhaoqiang Liu
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 wavelet low-frequency semantic compensation to FlowEdit | FlowEdit (FLUX) · PIE-Bench | PSNR: 22.23 -> 26.17 | Improved |
| add wavelet low-frequency semantic compensation to FlowEdit | FlowEdit (FLUX) · PIE-Bench | LPIPS x10^3: 111.32 -> 54.02 | Improved |
| add wavelet low-frequency semantic compensation to FlowEdit | FlowEdit (FLUX) · PIE-Bench | Structure distance x10^3: 27.61 -> 30.95 | Hurt |
| add wavelet low-frequency semantic compensation to FlowEdit | FlowEdit (FLUX) · PIE-Bench | CLIP edited: 22.62 -> 23.05 | Improved |
| compensate with full-spectrum signal instead of low-frequency only (ablation) | Ours (FLUX) · PIE-Bench | PSNR: 26.17 -> 23.61 | Hurt |
| compensate with high-frequency only instead of low-frequency (ablation) | Ours (FLUX) · PIE-Bench | CLIP edited: 23.05 -> 22.34 | Hurt |
| use Haar wavelet instead of Gaussian low-pass (ablation) | Ours (FLUX) · PIE-Bench | PSNR: 25.65 -> 26.17 | Improved |
| use Haar wavelet instead of Ideal FFT low-pass (ablation) | Ours (FLUX) · PIE-Bench | Structure distance x10^3: 37.48 -> 30.95 | Improved |
| increase wavelet level L from 1 to 4 (ablation) | Ours (FLUX) · PIE-Bench | PSNR: 25.26 -> 27.05 | Improved |
| increase semantic strength lambda from 2.0 to 4.0 (ablation) | Ours (FLUX) · PIE-Bench | PSNR: 26.88 -> 25.44 | Hurt |
| increase semantic strength lambda from 2.0 to 4.0 (ablation) | Ours (FLUX) · PIE-Bench | CLIP edited: 22.98 -> 22.94 | No clear effect |
| use proposed method instead of FlowAlign | FlowAlign (SD3) · PIE-Bench | PSNR: 23.94 -> 25.26 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 4 of 4 on most shared results.
- FlowEdit — FlowEdit: Inversion-Free Text-Based Editing Using Pre-Trained Flow Models (ICCV 2025) · better on 8 of 9 shared results
- DNA-Edit — DNAEdit: Direct Noise Alignment for Text-Guided Rectified Flow Editing (NeurIPS 2025) · better on 7 of 8 shared results
- FlowAlign — FlowAlign: Trajectory-Regularized, Inversion-Free Flow-based Image Editing (ICLR 2026) · better on 1 of 1 shared result
- FIA-Edit — FIA-Edit: Frequency-Interactive Attention for Efficient and High-Fidelity Inversion-Free Text-Guided Image Editing (AAAI 2026) · better on 1 of 1 shared result
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 0.89)
- Unveil Inversion and Invariance in Flow Transformer for Versatile Image Editing (CVPR 2025 · similarity 0.88)
- SplitFlow: Flow Decomposition for Inversion-Free Text-to-Image Editing (NeurIPS 2025 · similarity 0.87)
- Tuning-Free Inversion-Enhanced Control for Consistent Image Editing (AAAI 2024 · similarity 0.87)
- Object-Aware Inversion and Reassembly for Image Editing (ICLR 2024 · similarity 0.86)
- Semantic Image Inversion and Editing using Rectified Stochastic Differential Equations (ICLR 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.
