Research map / Image generation / Image restoration
Decouple to Reconstruct: High Quality UHD Restoration via Active Feature Disentanglement and Reversible Fusion
Yidi Liu, Dong Li, Yuxin Ma, Jie Huang, Wenlong Zhang, Xueyang Fu, Zheng-Jun Zha
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 Hi-CDL contrastive disentanglement to the VAE (ablation) | CD2-VAE · UHD-Blur deblurring | PSNR: 28.89 -> 29.48 | Improved |
| add OrthoGate to the VAE (ablation) | CD2-VAE · UHD-Blur deblurring | PSNR: 28.89 -> 29.32 | Improved |
| combine Hi-CDL and OrthoGate (ablation) | CD2-VAE · UHD-Blur deblurring | PSNR: 29.48 -> 29.84 | Improved |
| combine Hi-CDL and OrthoGate (ablation) | CD2-VAE · UHD-Blur deblurring | SSIM: 0.836 -> 0.861 | Improved |
| use reversible complex-domain CIMF-Net instead of non-reversible network (Set1) (ablation) | D2R-UHDNet · UHD-Blur deblurring | PSNR: 27.86 -> 29.84 | Improved |
| add cross-scale interaction via complex fusion instead of isolated INNs (Set2) (ablation) | D2R-UHDNet · UHD-Blur deblurring | PSNR: 29.26 -> 29.84 | Improved |
| use complex-domain fusion instead of cross-attention (Set3) (ablation) | D2R-UHDNet · UHD-Blur deblurring | Params (M): 4.289 -> 1.008 | Improved |
| keep complex numbers rather than concatenating real and imaginary parts (Set4) (ablation) | D2R-UHDNet · UHD-Blur deblurring | PSNR: 28.88 -> 29.84 | Improved |
| apply D2R framework to Restormer latent network (ablation) | Restormer · UHD-Blur deblurring | PSNR: 25.21 -> 29.78 | Improved |
| apply D2R framework to NAFNet latent network (ablation) | NAFNet · UHD-Blur deblurring | PSNR: 26.06 -> 29.88 | Improved |
| apply D2R framework to SFHformer latent network (ablation) | SFHformer · UHD-Blur deblurring | PSNR: 25.67 -> 29.84 | Improved |
| apply D2R framework to Restormer (ablation) | Restormer · UHD-Blur deblurring | Runtime (s): 8.8 -> 0.64 | Improved |
Closest papers by meaning
- Latent Harmony: Synergistic Unified UHD Image Restoration via Latent Space Regularization and Controllable Refinement (NeurIPS 2025 · similarity 0.91)
- DreamUHD: Frequency Enhanced Variational Autoencoder for Ultra-High-Definition Image Restoration (AAAI 2025 · similarity 0.91)
- UHD-processer: Unified UHD Image Restoration with Progressive Frequency Learning and Degradation-aware Prompts (CVPR 2025 · similarity 0.90)
- Correlation Matching Transformation Transformers for UHD Image Restoration (AAAI 2024 · similarity 0.88)
- From Zero to Detail: Deconstructing Ultra-High-Definition Image Restoration from Progressive Spectral Perspective (CVPR 2025 · similarity 0.87)
- SeNM-VAE: Semi-Supervised Noise Modeling with Hierarchical Variational Autoencoder (CVPR 2024 · similarity 0.84)
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.
