AI Paper Atlas

Research map / Image generation / Image restoration

Self-Adaptively Learning to Demoiré from Focused and Defocused Image Pairs

Lin Liu, Shanxin Yuan, Jianzhuang Liu, Liping Bao, Gregory Slabaugh, Qi Tian

NeurIPS 2020 · Image generation › Image restoration · approach: Diffusion priors for denoising and deblurring

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 changedWhereMeasuredEffect
Use the focused moiré image rather than a concatenation of focused and defocused images as D's input (ablation)FDNet · SynScreenMoirePSNR: 27.43 -> 31.77Improved
Use the focused moiré image rather than random noise as D's input (ablation)FDNet · SynScreenMoirePSNR: 28.42 -> 31.77Improved
Use the focused moiré image rather than only the defocused image as D's input (ablation)FDNet · SynScreenMoirePSNR: 27.17 -> 31.77Improved
Use end-to-end image prediction instead of predicting linear parameters for the defocused image (ablation)FDNet · SynScreenMoirePSNR: 29.57 -> 31.77Improved
Jointly optimize the demoiréing and blur-kernel networks instead of alternating their updates (ablation)FDNet · SynScreenMoirePSNR: 27.31 -> 31.77Improved
Use a blur-kernel generator network instead of a directly learnable kernel (ablation)FDNet · SynScreenMoirePSNR: 26.10 -> 31.77Improved
Use a U-Net rather than an encoder-decoder architecture (ablation)FDNet · SynScreenMoirePSNR: 0.11dBImproved
Use a U-Net rather than a ResNet architecture (ablation)FDNet · SynScreenMoirePSNR: 1.05dBImproved
Use FDNet rather than the strongest reported baselineDemoiréing methods · SynScreenMoirePSNR: 31.06 -> 31.77Improved
Use FDNet rather than the strongest reported baselineDemoiréing methods · SynTextureMoirePSNR: 24.70 -> 25.98Improved
Use FDNet rather than the strongest baselineDemoiréing methods · RealScreenMoireNIQE: 5.34 -> 5.11Improved
Use FDNet rather than the strongest baselineDemoiréing methods · RealTextureMoireBRISQUE: 42.13 -> 41.77Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy