AI Paper Atlas

Research map / Image generation / Super-resolution

Comparison of Full-Reference Image Quality Models for Optimization of Image Processing Systems

Keyan Ding, Kede Ma, Shiqi Wang, Eero P. Simoncelli

IJCV 2021 · Image generation › Super-resolution · approach: CNN and transformer SR architectures · 265 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
use DISTS or LPIPS instead of MAE as lossDNN denoiser · DenoisingSRCC vs subjective ranking: DISTS 0.346 vs MAE 0.527; LPIPS 0.709Mixed
use LPIPS instead of MAE as lossDNN deblurrer · DeblurringSRCC vs subjective ranking: 0.164 -> 0.855Improved
use DISTS instead of MAE as lossDNN super-resolution · Super-resolutionSRCC vs subjective ranking: 0.309 -> 0.782Improved
use DISTS instead of MAE as lossDNN compression · CompressionSRCC vs subjective ranking: 0.455 -> 0.855Improved
replace MAE with MS-SSIM as lossDNN image restoration (4 tasks) · Denoising, deblurring, super-resolution, compressionhuman rankingNo clear effect
replace LPIPS with DISTS as lossDNN image restoration · Super-resolutionhuman rankingImproved
replace VGG-based LPIPS loss with DISTS loss and fine-tune DeblurGAN-v2DeblurGAN-v2 · Köhler deblurringDISTS value: 0.22 -> 0.18Improved
replace original DeblurGAN-v2 loss with DISTS plus 0.001 adversarial lossDeblurGAN-v2 · Köhler deblurringvisual qualityImproved
pretrain network with MAE before IQA fine-tuningIQA-optimized DNNs · reference recovery and 4 tasksoptimization stabilityImproved
initialize recovery from JPEG image instead of white noise (ablation)surjective IQA models (GTI-CNN, DeepIQA, PieAPP) · Reference image recoveryrecovered structureImproved
use surjective mapping IQA models as objectiveGTI-CNN, DeepIQA, PieAPP · Reference image recoveryimage quality of resultHurt
use injective mapping IQA models as objectiveMS-SSIM, PAMSE, NLPD, DISTS · Reference image recoveryrecovery of referenceImproved

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