Research map / Image generation / Image restoration
Towards Flexible Blind JPEG Artifacts Removal
Jiaxi Jiang, Kai Zhang, Radu Timofte
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 |
|---|---|---|---|
| add dominant QF correction at inference (FBCNN-D) (ablation) | FBCNN · LIVE1 grayscale, non-aligned double JPEG, QF (10,30) | PSNR (dB): 28.29 -> 28.94 | Improved |
| add dominant QF correction at inference (FBCNN-D) (ablation) | FBCNN · LIVE1 grayscale, non-aligned double JPEG, QF (30,50) | PSNR (dB): 31.96 -> 32.31 | Improved |
| add double JPEG augmented training (FBCNN-A) on top of dominant QF correction (ablation) | FBCNN-D · LIVE1 grayscale, non-aligned double JPEG, QF (10,30) | PSNR (dB): 28.94 -> 29.38 | Improved |
| add double JPEG augmented training (FBCNN-A) (ablation) | FBCNN · LIVE1 grayscale, non-aligned double JPEG, QF (10,10) | PSNR (dB): 27.96 -> 28.25 | Improved |
| add double JPEG augmented training (FBCNN-A) (ablation) | FBCNN · LIVE1 grayscale, non-aligned double JPEG, QF (30,30) | PSNR (dB): 31.64 -> 31.94 | Improved |
| add double JPEG augmented training (FBCNN-A) (ablation) | FBCNN · LIVE1 grayscale, non-aligned double JPEG, QF (30,10) | PSNR (dB): 29.46 -> 29.44 | No clear effect |
| add double JPEG augmented training (FBCNN-A) (ablation) | FBCNN · LIVE1 grayscale, non-aligned double JPEG, QF (50,30) | PSNR (dB): 32.61 -> 32.69 | Improved |
| use FBCNN (QF predicted and embedded) instead of QGAC | FBCNN · Classic5 grayscale, single JPEG, QF 10 | PSNR (dB): 29.84 -> 30.12 | Improved |
| use FBCNN instead of QF-specific MWCNN | FBCNN · Classic5 grayscale, single JPEG, QF 10 | PSNR (dB): 30.01 -> 30.12 | Improved |
| train FBCNN on RGB (FBCNN-C) instead of QGAC | FBCNN-C · LIVE1 color, single JPEG, QF 10 | PSNR (dB): 27.62 -> 27.77 | Improved |
| blind single-JPEG models on non-aligned double JPEG with QF1 < QF2 (no correction) | FBCNN · LIVE1 grayscale, non-aligned double JPEG, QF (10,30) | PSNR (dB): 27.55 -> 28.29 | No clear effect |
Closest papers by meaning
- JPEG Artifacts Removal via Contrastive Representation Learning (ECCV 2022 · similarity 0.91)
- Compression-Aware One-Step Diffusion Model for JPEG Artifact Removal (ICCV 2025 · similarity 0.88)
- JPEG Artifacts Reduction via Deep Convolutional Sparse Coding (ICCV 2019 · similarity 0.88)
- DCTMamba: Advancing JPEG Image Restoration Through Long-Sequence Modeling and Adaptive Frequency Strategy (AAAI 2025 · similarity 0.84)
- JDEC: JPEG Decoding via Enhanced Continuous Cosine Coefficients (CVPR 2024 · similarity 0.83)
- DDR: Exploiting Deep Degradation Response as Flexible Image Descriptor (NeurIPS 2024 · similarity 0.83)
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.
