Research map / Image generation / Image restoration
Interpretable Unsupervised Diversity Denoising and Artefact Removal
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Replace the single-layer DivNoising architecture with a hierarchical VAE architecture | Unsupervised image denoising · 12 denoising datasets | PSNR | Helped |
| Use HDN rather than unsupervised denoising baselines | Unsupervised image denoising · 12 denoising datasets | PSNR | Helped |
| Use HDN rather than supervised baselines | Image denoising · Three occasions across the evaluated datasets | PSNR | Helped |
| Add batch normalization (ablation) | HDN · BSD68 | PSNR: about 0.5 dB | Helped |
| Add generative-path skip connections conditioning on higher layers (ablation) | HDN · BSD68 | PSNR: around 0.2 dB | Helped |
| Use free bits to stabilize training (ablation) | HDN · BSD68 | PSNR: around 0.15 dB | Helped |
| Increase the number of latent variables and gated residual blocks per layer (ablation) | HDN · Pixel-wise denoising datasets | PSNR | Mixed |
| Keep the full bottom-up input conditioning in all latent layers (ablation) | HDN structured-noise removal · Struct. Convallaria | PSNR | Hurt |
| Deactivate bottom-up conditioning in layer 1 (ablation) | HDN structured-noise removal · Struct. Convallaria | PSNR | Mixed |
| Deactivate bottom-up conditioning in layers 1 and 2 (ablation) | HDN structured-noise removal (HDN 3-6) · Struct. Convallaria | PSNR | Helped |
| Deactivate bottom-up conditioning in layers 1, 2, and 3 (ablation) | HDN structured-noise removal · Struct. Convallaria | PSNR | Mixed |
| Use HDN 3-6 rather than other unsupervised methods (ablation) | Structured-noise removal · Struct. Convallaria | PSNR | Helped |
Closest papers by meaning
- Fully Unsupervised Diversity Denoising with Convolutional Variational Autoencoders (ICLR 2021 · similarity 0.93)
- SeNM-VAE: Semi-Supervised Noise Modeling with Hierarchical Variational Autoencoder (CVPR 2024 · similarity 0.88)
- Denoising as Adaptation: Noise-Space Domain Adaptation for Image Restoration (ICLR 2025 · similarity 0.88)
- Resfusion: Denoising Diffusion Probabilistic Models for Image Restoration Based on Prior Residual Noise (NeurIPS 2024 · similarity 0.87)
- Residual Denoising Diffusion Models (CVPR 2024 · similarity 0.86)
- Stochastic Deep Restoration Priors for Imaging Inverse Problems (ICML 2025 · 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.