Research map / Image generation / Image restoration
Cascades of Regression Tree Fields for Image Restoration
Uwe Schmidt, Jeremy Jancsary, Sebastian Nowozin, Stefan Roth, Carsten Rother
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 cascade stages 1 to 5 (ablation) | RTF denoising · Gray denoising sigma=25 | PSNR: 28.24 -> 28.75 | Improved |
| add cascade stages 1 to 5 (quantized) (ablation) | RTF denoising · Gray denoising sigma=25 quantized | PSNR: 28.25 -> 28.74 | Improved |
| add cascade stages 1 to 6 (ablation) | RTF deblurring · Deblurring sigma=2.55 | PSNR: 26.33 -> 28.67 | Improved |
| add cascade stages 1 to 6 (ablation) | RTF deblurring · Deblurring sigma=7.65 | PSNR: 24.23 -> 25.89 | Improved |
| add stage 6 after stage 5 (ablation) | RTF deblurring · Deblurring sigma=2.55 | PSNR: 28.65 -> 28.67 | No clear effect |
| train color model with channels jointly instead of channel-independent gray cascade (ablation) | RTF · Color denoising | PSNR: 28.68 -> 30.57 | Improved |
| add second stage to color joint model (ablation) | RTF color · Color denoising | PSNR: 30.01 -> 30.57 | Improved |
| train with GT kernels, test with estimated kernels, add stage 2 (ablation) | RTF · Levin deblurring, Xu-Jia kernels | PSNR: 29.04 -> 28.29 | Hurt |
| train with GT + noisy kernels instead of GT kernels (ablation) | RTF1 · Levin deblurring, Xu-Jia kernels | PSNR: 29.04 -> 30.25 | Improved |
| add stage 2 when trained on noisy GT kernels (ablation) | RTF · Levin deblurring, Xu-Jia kernels | PSNR: 30.25 -> 29.56 | Hurt |
| train with mix of GT and estimated kernels instead of GT kernels (ablation) | RTF2 · Levin deblurring, Xu-Jia kernels | PSNR: 28.29 -> 30.84 | Improved |
| add stage 2 when trained on mix of GT and estimated kernels (ablation) | RTF · Levin deblurring, Levin kernel estimates | PSNR: 29.90 -> 30.40 | Improved |
Closest papers by meaning
- Unsupervised Image Restoration Using Partially Linear Denoisers (TPAMI 2021 · similarity 0.88)
- Learning Data Terms for Non-blind Deblurring (ECCV 2018 · similarity 0.87)
- Deep Gaussian Conditional Random Field Network: A Model-Based Deep Network for Discriminative Denoising (CVPR 2016 · similarity 0.87)
- Denoising as Adaptation: Noise-Space Domain Adaptation for Image Restoration (ICLR 2025 · similarity 0.87)
- DeblurDiff: Real-Word Image Deblurring with Generative Diffusion Models (NeurIPS 2025 · similarity 0.86)
- A Modular Conditional Diffusion Framework for Image Reconstruction (NeurIPS 2024 · similarity 0.85)
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.
