Research map / Image generation / Image restoration
Learning Dual Priors for JPEG Compression Artifacts Removal
Xueyang Fu, Xi Wang, Aiping Liu, Junwei Han, Zheng-Jun Zha
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 changed | Where | Measured | Effect |
|---|---|---|---|
| replace the l1 sparsity proximal operator with the learned multi-scale non-local dilated de-blocker f_p (ablation) | dual-prior unfolding network · BSD500 QF=10 | PSNR: 28.18 -> 29.48 | Improved |
| replace DnCNN de-blocker with the proposed non-local dilated de-blocker f_p (ablation) | dual-prior unfolding network · BSD500 QF=10 | PSNR-B: 27.95 -> 29.13 | Improved |
| replace DnCNN de-blocker with the proposed de-blocker f_p (ablation) | dual-prior unfolding network · BSD500 QF=10 | PSNR: 28.97 -> 29.48 | Improved |
| add the semantic-level classifier prior f_s on the gradient (ablation) | pixel-level prior only network · visual example | edge sharpness (visual) | Improved |
| increase number of unfolding stages from K=1 to K=3 (ablation) | dual-prior unfolding network · BSD500 QF=10 | PSNR/SSIM | Improved |
| increase number of unfolding stages from K=5 to K=7 (ablation) | dual-prior unfolding network · BSD500 QF=10 | PSNR/SSIM | Hurt |
| use the proposed dual-prior network instead of RDN | JPEG de-blocking · Classic5 QF=10 | PSNR: 30.03 -> 30.26 | Improved |
| use the proposed dual-prior network instead of RDN | JPEG de-blocking · Twitter | PSNR: 27.44 -> 28.27 | Improved |
| use the proposed network instead of RDN on LIVE1 QF=30 | JPEG de-blocking · LIVE1 QF=30 | PSNR: 33.54 -> 33.43 | Hurt |
Closest papers by meaning
- JPEG Artifacts Removal via Contrastive Representation Learning (ECCV 2022 · similarity 0.91)
- JPEG Artifacts Reduction via Deep Convolutional Sparse Coding (ICCV 2019 · similarity 0.90)
- Compression-Aware One-Step Diffusion Model for JPEG Artifact Removal (ICCV 2025 · similarity 0.90)
- Deep Residual Learning in the JPEG Transform Domain (ICCV 2019 · similarity 0.88)
- D3: Deep Dual-Domain Based Fast Restoration of JPEG-Compressed Images (CVPR 2016 · similarity 0.88)
- DDR: Exploiting Deep Degradation Response as Flexible Image Descriptor (NeurIPS 2024 · similarity 0.86)
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.
