Research map / Image generation / Super-resolution
Faithful Extreme Rescaling via Generative Prior Reciprocated Invertible Representations
Zhixuan Zhong, Liangyu Chai, Yang Zhou, Bailin Deng, Jia Pan, Shengfeng He
Check 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 |
|---|---|---|---|
| combine invertible rescaling with a reciprocated generative prior | GRAIN · CelebA-HQ test | LPIPS: 0.3891 -> 0.2686 | Improved |
| combine invertible rescaling with a reciprocated generative prior | GRAIN · CelebA-HQ test | PSNR: 25.15 -> 22.30 | Hurt |
| add the generative prior to invertible-only rescaling (ablation) | GRAIN · CelebA-HQ test | LPIPS: 0.3379 -> 0.2686 | Improved |
| replace the pSp encoder with scale-specific latent-code prediction (ablation) | GRAIN · CelebA-HQ test | LPIPS: 0.4456 -> 0.2686 | Improved |
| add the invertible encoder that embeds HR information in the LR image (ablation) | GRAIN · CelebA-HQ test | LPIPS: 0.3299 -> 0.2686 | Improved |
| replace image-level fusion with multi-scale feature decoding (ablation) | GRAIN · CelebA-HQ test | LPIPS: 0.2843 -> 0.2686 | Improved |
| increase LR resolution from 16 × 16 to 128 × 128 (ablation) | GRAIN · CelebA-HQ test | LPIPS: 0.2686 -> 0.0719 | Improved |
| use GRAIN rescaling rather than JPEG compression | GRAIN · CelebA-HQ test, comparable PSNR | storage size: 724 vs 18803 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- GLEAN (64×) — GLEAN: Generative Latent Bank for Large-Factor Image Super-Resolution (CVPR 2021) · better on 3 of 3 shared results
- PULSE (64×) — PULSE: Self-Supervised Photo Upsampling via Latent Space Exploration of Generative Models (CVPR 2020) · better on 3 of 3 shared results
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
- Timestep-Aware Diffusion Model for Extreme Image Rescaling (ICCV 2025 · similarity 0.90)
- Invertible generative models for inverse problems: mitigating representation error and dataset bias (ICML 2020 · similarity 0.89)
- Plug-and-Play Tri-Branch Invertible Block for Image Rescaling (AAAI 2025 · similarity 0.89)
- Self-Asymmetric Invertible Network for Compression-Aware Image Rescaling (AAAI 2023 · similarity 0.89)
- When Latents Forget Pixels: Restoring Fidelity in Diffusion Transformer Super-Resolution (NeurIPS 2026 · similarity 0.87)
- Noise-Started One-Step Image Super-Resolution via LR-Conditioned SplitMeanFlow and GAN Refinement (NeurIPS 2026 · 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.
