AI Paper Atlas

Research map / Image generation / Image restoration

Microscopy Image Restoration with Deep Wiener-Kolmogorov Filters

Valeriya Pronina, Filippos Kokkinos, Dmitry V. Dylov, Stamatios Lefkimmiatis

ECCV 2020 · Image generation › Image restoration · approach: Diffusion-based universal restoration · 31 citations (OpenAlex)

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
predict regularization kernels per image with a KPN instead of fixed learnable kernels (ablation)Wiener filter (WF-K) · Gaussian deblurring (noise std 0.001)PSNR: 35.66 -> 38.72Improved
predict kernels per pixel (spatially adaptive) instead of per image (ablation)WF-KPN · Gaussian deblurring (noise std 0.1)PSNR: 29.20 -> 32.60Improved
use WF-KPN-SA instead of UNet baselineUNet · Gaussian deblurring (noise std 0.1)PSNR: 32.40 -> 32.60Improved
combine Wiener filter with UNet regularizer gradient (WF-UNet) instead of plain UNetUNet · Gaussian deblurring (noise std 0.1)PSNR: 32.40 -> 32.01Hurt
combine Wiener filter with UNet regularizer gradient (WF-UNet) instead of plain UNetUNet · Poisson deblurring (peak 1)PSNR: 28.50 -> 29.04Improved
increase kernels from 8 of size 3 to 24 of size 5 (ablation)WF-K · Gaussian deblurringPSNR: 35.66 -> 35.81 at std 0.001; 29.27 -> 29.21 at std 0.1Mixed
increase kernels from 8 of size 3 to 24 of size 5 (ablation)WF-KPN · Gaussian deblurring (noise std 0.001)PSNR: 38.72 -> 39.95Improved
increase kernels from 8 of size 3 to 24 of size 5 (ablation)WF-KPN · Gaussian deblurring (noise std 0.1)PSNR: 29.20 -> 29.20No clear effect
use noise-blind WF-KPN-SA versus noise-aware DMSP-NA at very low noiseWF-KPN-SA · Gaussian deblurring (noise std 0.001)PSNR: 40.44 -> 39.86Hurt
use spatially adaptive kernels instead of per-image kernels (ablation)WF-KPN · Runtime 256x256 image (GPU)time (ms): 7.7 -> 341.4Hurt

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