Research map / Image generation / Super-resolution
Deep Decomposition Learning for Inverse Imaging Problems
Dongdong Chen, Mike E. Davies
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 range-space projector Pr to DDN (versus none) (ablation) | DDN · Set5 SR x2 (sigma=0.1, BSDS300 train) | PSNR (dB): 22.31 -> 25.82 | Improved |
| add null-space projector Pn to DDN (versus none) (ablation) | DDN · Set5 SR x2 (sigma=0.1, BSDS300 train) | PSNR (dB): 22.31 -> 20.95 for Pn alone | Improved |
| add Pn on top of Pr (ablation) | DDN · Set5 SR x2 (sigma=0.1, BSDS300 train) | PSNR (dB): 25.82 -> 26.30 | Improved |
| add Pr on top of Pn (ablation) | DDN · Set5 SR x2 (sigma=0.1, BSDS300 train) | PSNR (dB): 20.95 -> 26.30 | Improved |
| add Pr and Pn physics to residual learning (DDN1) in generalization (ablation) | DDN · Set5 SR x2 (sigma=0.1, BSDS300 train) | generalization error x10^-3: 1.75 -> 1.26 | Mixed |
| use cascade instead of independent connection (ablation) | DDN · Set5 SR x2 (sigma=0.1, BSDS300 train) | PSNR (dB): 26.15 -> 26.30 | Improved |
| use decoupled instead of joint training in cascade DDN (ablation) | DDN · Set5 SR x2 (sigma=0.1, BSDS300 train) | PSNR (dB): 26.30 -> 26.38 | No clear effect |
| use DDN instead of ResUnet | CS-MRF · CS-MRF X reconstruction (N=10) | NMSE x10^-2: 1.119 -> 1.056 | Improved |
| use DDN instead of NPGD with few samples | CS-MRF · CS-MRF X reconstruction (N=10) | NMSE x10^-2: 1.689 -> 1.056 | Improved |
| use DDN instead of NSN | CS-MRF · CS-MRF T1 map (N=100) | PSNR (dB): 18.49 -> 18.92 | Improved |
| use DDN instead of NSN | CS-MRF · CS-MRF T1 map (N=10) | PSNR (dB): 15.25 -> 15.90 | Improved |
| use DDN instead of NPGD for inference cost | CS-MRF · CS-MRF reconstruction | inference time (sec): 0.15 -> 0.06 | Improved |
Closest papers by meaning
- NPN: Non-Linear Projections of the Null-Space for Imaging Inverse Problems (NeurIPS 2025 · similarity 0.90)
- An analytic theory of convolutional neural network inverse problems solvers (ICML 2026 · similarity 0.89)
- Learning Proximal Operators: Using Denoising Networks for Regularizing Inverse Imaging Problems (ICCV 2017 · similarity 0.88)
- An inner-loop free solution to inverse problems using deep neural networks (NeurIPS 2017 · similarity 0.87)
- A Restoration Network as an Implicit Prior (ICLR 2024 · similarity 0.87)
- Untrained Neural Nets for Snapshot Compressive Imaging: Theory and Algorithms (NeurIPS 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.
