Research map / Image generation / Image restoration
Decouple Learning for Parameterized Image Operators
Qingnan Fan, Dongdong Chen, Lu Yuan, Gang Hua, Nenghai Yu, Baoquan Chen
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 |
|---|---|---|---|
| train one network jointly on continuous random parameter values instead of one per value | L0 filter network · L0 smoothing | PSNR diff: 0.86 | Hurt |
| train jointly on random parameters instead of per-value networks | WLS filter network · WLS filter | PSNR diff: 0.61 | Hurt |
| train jointly on random parameters instead of per-value networks | RTV filter network · RTV filter | PSNR diff: 0.31 | Hurt |
| train jointly on random parameters instead of per-value networks | Super resolution network · Super resolution | PSNR: 29.29 -> 29.23 | No clear effect |
| train jointly on random parameters instead of per-value networks | Denoising network · Denoising | PSNR: 28.77 -> 28.73 | No clear effect |
| train jointly on random parameters instead of per-value networks | Deblock network · Deblock | PSNR: 30.37 -> 30.30 | No clear effect |
| use two-dimensional parameters (lambda, sigma) in one network, varying sigma | RTV network · RTV (lambda=0.01) | PSNR: 40.86 -> 40.93 | No clear effect |
| use two-dimensional parameters (lambda, sigma) in one network, varying lambda | RTV network · RTV (sigma=3) | PSNR: 40.86 -> 40.14 | Hurt |
| joint training over 6 filters or 4 restoration tasks instead of one operator (ablation) | Decouple learning · 10 image operators | PSNR (avg): 34.86 -> 33.49 | Hurt |
| joint training over all 10 operators instead of one operator (ablation) | Decouple learning · 10 image operators | PSNR (avg): 34.86 -> 32.89 | Hurt |
| joint training over all 10 operators instead of one operator (ablation) | Decouple learning · 10 image operators | SSIM (avg): 0.925 -> 0.914 | Hurt |
| disable parameter specification for restoration tasks in joint training | Decouple learning · restoration tasks | PSNR | No clear effect |
Closest papers by meaning
- A General Decoupled Learning Framework for Parameterized Image Operators (TPAMI 2019 · similarity 0.97)
- Decoupled Networks (CVPR 2018 · similarity 0.85)
- Fast Image Processing With Fully-Convolutional Networks (ICCV 2017 · similarity 0.84)
- Multilinear Operator Networks (ICLR 2024 · similarity 0.83)
- MgNO: Efficient Parameterization of Linear Operators via Multigrid (ICLR 2024 · similarity 0.83)
- What's in a Prior? Learned Proximal Networks for Inverse Problems (ICLR 2024 · similarity 0.83)
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.
