Research map / Image generation / Image restoration
FideDiff: Efficient Diffusion Model for High-Fidelity Image Motion Deblurring
Xiaoyang Liu, Zhengyan Zhou, Zihang Xu, Jiezhang Cao, Zheng Chen, Yulun 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 |
|---|---|---|---|
| reduce VAE downsampling from d=8 to d=4 (patch 384) (ablation) | FideDiff foundation model · GoPro | PSNR: 26.26 -> 27.77 | Improved |
| increase training patch size from 384 to 512 (d=4) (ablation) | FideDiff foundation model · GoPro | PSNR: 27.77 -> 28.51 | Improved |
| use LPIPS loss instead of DISTS loss (ablation) | FideDiff foundation model · GoPro | LPIPS: 0.1288 -> 0.1093 | Improved |
| add GAN discriminator (ablation) | FideDiff foundation model · GoPro | DISTS: 0.0732 -> 0.0633 | Improved |
| add learnable text embedding (ablation) | FideDiff foundation model · GoPro | LPIPS: 0.1138 -> 0.1093 | Improved |
| add edge-enhanced perceptual loss (EA-LPIPS) (ablation) | FideDiff foundation model · GoPro | LPIPS: 0.1119 -> 0.1093 | Improved |
| add Kernel ControlNet (ablation) | FideDiff foundation model · GoPro/HIDE/RealBlur | full-reference metrics | Improved |
| add time-consistency training (CT) (ablation) | FideDiff · GoPro/HIDE/RealBlur-R | LPIPS/DISTS | Improved |
| add t-prediction to the CT model (ablation) | FideDiff · RealBlur | full-reference metrics | Improved |
| add t-prediction to the CT model (ablation) | FideDiff · GoPro/HIDE | full-reference metrics | No clear effect |
| add motion alignment module based on MISCFilter instead of Kernel ControlNet (ablation) | FideDiff · GoPro | full-reference metrics | Hurt |
Closest papers by meaning
- DeblurDiff: Real-Word Image Deblurring with Generative Diffusion Models (NeurIPS 2025 · similarity 0.91)
- BlurDM: A Blur Diffusion Model for Image Deblurring (NeurIPS 2025 · similarity 0.90)
- Improving Diffusion-Based Image Restoration with Error Contraction and Error Correction (AAAI 2024 · similarity 0.89)
- A Modular Conditional Diffusion Framework for Image Reconstruction (NeurIPS 2024 · similarity 0.88)
- Blind Image Restoration via Fast Diffusion Inversion (NeurIPS 2024 · similarity 0.88)
- Denoising as Adaptation: Noise-Space Domain Adaptation for Image Restoration (ICLR 2025 · similarity 0.88)
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.
