AI Paper Atlas

Research map / Image generation / Image restoration

ExposureDiffusion: Learning to Expose for Low-light Image Enhancement

Yufei Wang, Yi Yu, Wenhan Yang, Lanqing Guo, Lap-Pui Chau, Alex C. Kot, Bihan Wen

ICCV 2023 · Image generation › Image restoration · approach: Diffusion-based universal restoration · 109 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
train with the exposure-diffusion process instead of a feed-forward mapping, same P+G noise model and backboneUNet · SID Sony x250PSNR: 34.39 -> 36.02Improved
apply ExposureDiffusion to small NAFNet-1 backboneNAFNet-1 · SID x100 (P+G)PSNR: 32.37 -> 36.74Improved
apply ExposureDiffusion on top of SOTA PMN noise modelPMN noise model · SID x100PSNR: 39.77 -> 39.80No clear effect
apply ExposureDiffusion on ELD noise modelUNet · SID x100PSNR: 39.27 -> 39.37Improved
add adaptive residual layer (ablation)ExposureDiffusion · SID example imagePSNR: 32.57 -> 35.44Improved
use ExposureDiffusion for unseen amplifying ratios (train x100, test x300)PMN + NAFNet-2 · SID x300 OODPSNR: 36.31 -> 36.83Improved
increase inference steps to 10 for baseline (ablation)feed-forward baseline · SIDPSNRHurt
replace vanilla conditional diffusion with exposure diffusion at similar model size (ablation)UNet-sized backbone · SID x100PSNR: 33.47 -> 38.88Improved
use L1 loss instead of KL-divergence loss (ablation)ExposureDiffusion · SIDconverged performanceNo clear effect
small ExposureDiffusion model vs larger feed-forward modelNAFNet-1 vs NAFNet-2 · SIDPSNR and cost: about 25% of parameters and FLOPsImproved
apply ExposureDiffusion on Canon EOS700D x200 vs P+GUNet · ELD Canon EOS700D x200PSNR: 37.86 -> 37.71Hurt

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