AI Paper Atlas

Research map / Image generation / Super-resolution

Kernel Modeling Super-Resolution on Real Low-Resolution Images

Ruofan Zhou, Sabine Susstrunk

ICCV 2019 · Image generation › Super-resolution · approach: Blind super-resolution · 170 citations (OpenAlex)

Read the paper

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
add realistic blur-kernels estimated from real photographs to training (ablation)KMSR · DIV2K val, Gaussian g1.6 x2PSNR (dB): 25.62 -> 27.63Improved
add realistic blur-kernels estimated from real photographs to training (ablation)KMSR · DIV2K val, realistic kernels x2PSNR (dB): 25.29 -> 27.52Improved
add realistic blur-kernels estimated from real photographs to training (ablation)KMSR · DIV2K val, Gaussian g1.25 x2PSNR (dB): 26.47 -> 27.94Improved
add GAN-generated kernels to the estimated kernel pool (ablation)KMSR · DIV2K val, Gaussian g1.6 x2PSNR (dB): 27.02 -> 27.63Improved
add GAN-generated kernels to the estimated kernel pool (ablation)KMSR · DIV2K val, Gaussian g1.25 x2PSNR (dB): 27.42 -> 27.94Improved
add GAN-generated kernels to the estimated kernel pool (ablation)KMSR · DIV2K val, realistic kernels x2PSNR (dB): 27.10 -> 27.52Improved
add GAN-generated kernels to the estimated kernel pool (ablation)KMSR · DIV2K val, Gaussian g1.7 x2PSNR (dB): 27.90 -> 27.15Hurt
add GAN-generated kernels to the estimated kernel pool (ablation)KMSR · DIV2K val, bicubic kernel x2PSNR (dB): 33.28 -> 33.52Hurt
add GAN kernel augmentation versus simple geometric augmentation, averagedKMSR · DIV2K val x2, average over kernelsPSNR (dB): 0.5dBImproved
train on realistic kernel pool instead of bicubic-only trainingKMSR vs DBPN · DIV2K val, bicubic kernel x2PSNR (dB): 33.84 -> 33.52Hurt
train on realistic kernel pool instead of bicubic-only trainingKMSR vs DBPN · DIV2K val, Gaussian g1.6 x2PSNR (dB): 25.70 -> 27.63Improved
train on realistic kernel pool instead of bicubic-only trainingKMSR vs DBPN · x2 zoom-in datasetPSNR (dB): 27.42 -> 29.13Improved

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