Research map / Image generation / Super-resolution
Hierarchical Conditional Flow: A Unified Framework for Image Super-Resolution and Image Rescaling
Jingyun Liang, Andreas Lugmayr, Kai Zhang, Martin Danelljan, Luc Van Gool, Radu Timofte
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 conditional priors y2 and y1 (hierarchical conditioning) (ablation) | HCFlow · DIV2K val x4 SR | PSNR: 18.95 -> 28.71 | Improved |
| use hierarchical instead of single-scale conditional mechanism (ablation) | HCFlow · DIV2K val x4 SR | PSNR: 28.60 -> 28.71 | Improved |
| use hierarchical instead of single-scale conditional mechanism (ablation) | HCFlow · DIV2K val x4 SR | LPIPS (tau=0.9): 0.126 -> 0.124 | Improved |
| add LR image space y2 to the latent space (ablation) | HCFlow without conditional priors · DIV2K val x4 SR | PSNR: 4.76 -> 18.95 | Improved |
| use external ground-truth LR as prior instead of LR in latent space (ablation) | HCFlow · DIV2K val x4 SR | PSNR: 28.71 -> 28.73 | No clear effect |
| condition on ground-truth LR y* additionally to generated y2 (ablation) | HCFlow · DIV2K val x4 SR | PSNR: 28.71 -> 28.71 | No clear effect |
| add HR pixel loss (HCFlow+) | HCFlow · DIV2K val x4 SR | PSNR: 28.71 -> 29.25 | Improved |
| add perceptual and GAN loss (HCFlow++) | HCFlow · DIV2K val x4 SR | LPIPS: 0.124 -> 0.110 | Improved |
| raise sampling temperature from 0 to 0.9 | HCFlow · DIV2K val x4 SR | LPIPS: 0.285 -> 0.124 | Improved |
| raise sampling temperature from 0 to 0.9 | HCFlow · DIV2K val x4 SR | PSNR: 28.71 -> 27.02 | Hurt |
| use HCFlow instead of IRN for rescaling | IRN · Set14 image rescaling x4 | PSNR: 32.67 -> 33.02 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- RankSRGAN — RankSRGAN: Super Resolution Generative Adversarial Networks With Learning to Rank (TPAMI 2021) · 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
- SRFlow: Learning the Super-Resolution Space with Normalizing Flow (ECCV 2020 · similarity 0.91)
- Boosting Flow-based Generative Super-Resolution Models via Learned Prior (CVPR 2024 · similarity 0.90)
- STARFlow: Scaling Latent Normalizing Flows for High-resolution Image Synthesis (NeurIPS 2025 · similarity 0.88)
- HiFlow: Training-free High-Resolution Image Generation with Flow-Aligned Guidance (NeurIPS 2025 · similarity 0.87)
- Posterior-Mean Rectified Flow: Towards Minimum MSE Photo-Realistic Image Restoration (ICLR 2025 · similarity 0.86)
- Efficient Conditional Diffusion Model with Probability Flow Sampling for Image Super-resolution (AAAI 2024 · 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.
