Research map / Image generation / Image restoration
When Fast Fourier Transform Meets Transformer for Image Restoration
Xingyu Jiang, Xiuhui Zhang, Ning Gao, Yue Deng
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 spatial branch to the frequency branch (hybrid structure) (ablation) | SFHformer LGPM · Rain200L | PSNR: 41.71 -> 41.85 | Improved |
| add frequency branch to the spatial branch (hybrid structure) (ablation) | SFHformer LGPM · Rain200L | PSNR: 41.62 -> 41.85 | Improved |
| add frequency-conditional positional encoding (FCPE) (ablation) | SFHformer · Rain200L | PSNR: -0.05 dB when removed | Improved |
| use frequency dynamic convolution instead of static PW conv (ablation) | SFHformer · Rain200L | PSNR: -0.11 dB when replaced | Improved |
| use Multi-kernel ConvFFN instead of DFN, GDFN or MSFN (ablation) | SFHformer · Rain200L | PSNR: +0.06 dB (DFN), +0.12 dB (GDFN), +0.03 dB (MSFN) | Improved |
| use batch norm instead of layer norm (ablation) | SFHformer · Rain200L | PSNR: +0.07 dB | Improved |
| use dual-domain (spatial plus FFT) loss instead of single L1 or FFT loss (ablation) | SFHformer · Rain200L | PSNR | Improved |
| use dual-branch spatial module with DW and dilated conv instead of single branch (ablation) | SFHformer spatial branch · Rain200L | PSNR: 41.76 -> 41.85 | Improved |
| replace self-attention with FFT-based token mixer | transformer image restoration · SOTS-outdoor dehazing | PSNR and FLOPs: 38.09 -> 38.83 PSNR; 88.1G -> 26.59G FLOPs vs MB-TaylorFormer | Improved |
| SFHformer versus Retinexformer on LOL-v1 | SFHformer · LOL-v1 | PSNR: 25.16 -> 24.29 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 3 on most shared results.
- Retinexformer — Retinexformer: One-stage Retinex-based Transformer for Low-light Image Enhancement (ICCV 2023) · better on 1 of 2 shared results
- IRNeXt — IRNeXt: Rethinking Convolutional Network Design for Image Restoration (ICML 2023) · better on 1 of 1 shared result
- Restormer — Restormer: Efficient Transformer for High-Resolution Image Restoration (CVPR 2022) · 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
- Image Restoration via Multi-Domain Learning (TPAMI 2026 · similarity 0.91)
- Improving the Learning Capability of Small-size Image Restoration Network by Deep Fourier Shifting (NeurIPS 2024 · similarity 0.88)
- Adapt or Perish: Adaptive Sparse Transformer with Attentive Feature Refinement for Image Restoration (CVPR 2024 · similarity 0.87)
- AdaIR: Adaptive All-in-One Image Restoration via Frequency Mining and Modulation (ICLR 2025 · similarity 0.86)
- DreamUHD: Frequency Enhanced Variational Autoencoder for Ultra-High-Definition Image Restoration (AAAI 2025 · similarity 0.86)
- From Zero to Detail: Deconstructing Ultra-High-Definition Image Restoration from Progressive Spectral Perspective (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.
