AI Paper Atlas

Research map / Image generation / Image restoration

Efficient Frequency-Domain Image Deraining with Contrastive Regularization

Ning Gao, Xingyu Jiang, Xiuhui Zhang, Yue Deng

ECCV 2024 · Image generation › Image restoration · approach: Dehazing and deraining · 86 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 FCR to training (ablation)FADformer · Rain200HPSNR (dB): 30.25 -> 30.73Improved
add FCR to training (ablation)FADformer · DDN-DataPSNR (dB): 33.51 -> 33.61Improved
add FCR to training (ablation)MPRNet · Rain200HPSNR (dB): 30.32 -> 30.62Improved
add FCR to dehazing training (ablation)FFA-Net · SOTS-IndoorPSNR (dB): 36.39 -> 37.29Improved
use frequency-domain contrastive regularization instead of VGG perceptual contrastive loss (ablation)FADformer · Rain200HPSNR (dB): 30.53 -> 30.73Improved
include negative samples in frequency regularization (ablation)FADformer · Rain200HPSNR (dB): 30.48 -> 30.73Improved
use too many negative samples (1:6 instead of 1:2) (ablation)FCR · Rain200HPSNR (dB): 30.73 -> 30.67Hurt
replace self-attention (MDTA) token mixer with FFCM (ablation)FADformer · Rain200HPSNR (dB): 29.66 -> 29.95 at depth {2,3,4,3,2}Improved
add FFT/IFFT frequency branch to FFCM (ablation)FFCM · Rain200HPSNR (dB): 29.27 -> 30.73Improved
replace PGFN with plain DFN (no prior) (ablation)FADformer · Rain200HPSNR (dB): 30.34 -> 30.73Improved
introduce RCP by concatenation instead of gating (ablation)PGFN · Rain200HPSNR (dB): 30.73 -> 29.52Hurt
use FADformer instead of DRSformerFADformer · SPA-DataPSNR (dB): 48.54 -> 49.21Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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