AI Paper Atlas

Research map / Image generation / Image restoration

Adaptive Denoising via GainTuning

Sreyas Mohan, Joshua L. Vincent, Ramon Manzorro, Peter Crozier, Carlos Fernandez-Granda, Eero P Simoncelli

NeurIPS 2021 · Image generation › Image restoration · approach: Self-supervised image denoising · 10 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
add GainTuning (per-channel gain adaptation) to pre-trained denoiserBFCNN · Set12, sigma=30PSNR (dB): 29.52 -> 29.61Improved
add GainTuning to pre-trained denoiser on out-of-distribution noiseDnCNN · Set12, sigma=80PSNR (dB): 18.48 -> 24.57Improved
add GainTuning to pre-trained denoiser on out-of-distribution noiseDnCNN · BSD68, sigma=70PSNR (dB): 22.15 -> 24.89Improved
GainTuning instead of LIDIA adaptationDnCNN · Set12, sigma=80PSNR (dB): 24.17 -> 24.57Improved
GainTuning versus Self2Self trained only on test imageDnCNN · Set12, sigma=70PSNR (dB): 24.61 -> 25.54Improved
GainTuning versus supervised model trained on wide noise rangeDnCNN · Set12, sigma=70PSNR (dB): 25.59 vs 25.54No clear effect
use SURE loss instead of blind-spot loss in GainTuning (ablation)GainTuning · Gaussian noise denoisingPSNRImproved
tune all parameters instead of only channel gains (ablation)pre-trained CNN denoiser · electron microscopy (TEM) datadenoising qualityHurt
tune all parameters instead of only gains (ablation)pre-trained CNN denoiser · Urban100 (high self-similarity)PSNRImproved
Self2Self trained only on test data instead of GainTuningout-of-distribution image content · piecewise constant to natural imagesPSNRImproved
add GainTuning to well-optimized pre-trained denoiser on in-distribution imagesDnCNN with SURE · BSD68 in-distributionPSNR: max degradation about 0.02 dB in 3 of 408 evaluationsMixed

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