Research map / Image generation / Image restoration
Timestep-Aware Diffusion Model for Extreme Image Rescaling
Ce Wang, Zhenyu Hu, Wanjie Sun, Zhenzhong Chen
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 rescaling in latent space (instead of pixel space) (ablation) | TADM · DIV2K 16x rescaling | LPIPS: 0.3630 -> 0.2979 | Improved |
| add rescaling in latent space (instead of pixel space) (ablation) | TADM · DIV2K 16x rescaling | DISTS: 0.1154 -> 0.0886 | Improved |
| add SD prior perceptual enhancement (ablation) | TADM · DIV2K 16x rescaling | LPIPS: 0.4675 -> 0.2979 | Improved |
| add SD prior perceptual enhancement (ablation) | TADM · DIV2K 16x rescaling | DISTS: 0.3109 -> 0.0886 | Improved |
| add time-step alignment (TPM and time scheduler) instead of fixed time-step 1 or 999 (ablation) | TADM · DIV2K 16x rescaling | fidelity and perceptual quality | Improved |
| add INN domain converter in DFRM (ablation) | DFRM · DIV2K 16x rescaling | validation reconstruction loss | Improved |
| add pixel guidance module in DFRM (ablation) | DFRM · DIV2K 16x rescaling | reconstruction loss | No clear effect |
| use patch size 96 instead of too large or small patches in tiled inference (ablation) | TADM tiled inference · DIV2K 16x rescaling | performance | Improved |
| replace VQIR with TADM at 32x | VQIR · DIV2K 32x rescaling | DISTS: 0.2663 -> 0.1684 | Improved |
| replace HCFlow with TADM at 16x on Urban100 | HCFlow · Urban100 16x rescaling | SSIM: 0.6335 -> 0.6495 | Improved |
| replace HCFlow with TADM at 16x on DIV2K | HCFlow · DIV2K 16x rescaling | PSNR: 26.66 -> 23.98 | Hurt |
Closest papers by meaning
- Simpler Diffusion: 1.5 FID on ImageNet512 with Pixel-space Diffusion (CVPR 2025 · similarity 0.89)
- FAM Diffusion: Frequency and Attention Modulation for High-Resolution Image Generation with Stable Diffusion (CVPR 2025 · similarity 0.89)
- Learning Diffusion Texture Priors for Image Restoration (CVPR 2024 · similarity 0.88)
- RepLDM: Reprogramming Pretrained Latent Diffusion Models for High-Quality, High-Efficiency, High-Resolution Image Generation (NeurIPS 2025 · similarity 0.88)
- Taming Diffusion Prior for Image Super-Resolution with Domain Shift SDEs (NeurIPS 2024 · similarity 0.88)
- One-Step Effective Diffusion Network for Real-World Image Super-Resolution (NeurIPS 2024 · 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.
