Research map / Image generation / Image restoration
Quantization Guided JPEG Artifact Correction
Max Ehrlich, Ser-Nam Lim, Larry Davis, Abhinav Shrivastava
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 |
|---|---|---|---|
| use a single quantization-guided model instead of IDCN's mismatched quality-10 model on quality-50 images | Quantization Guided JPEG Artifact Correction · Live-1 generalization (image quality 50) | PSNR: 30.19 -> 32.78 | Improved |
| use a single quantization-guided model instead of IDCN's mismatched quality-20 model on quality-50 images | Quantization Guided JPEG Artifact Correction · Live-1 generalization (image quality 50) | PSNR: 31.91 -> 32.78 | Improved |
| use a single quantization-guided model instead of IDCN's quality-10 model on quality-20 images | Quantization Guided JPEG Artifact Correction · Live-1 generalization (image quality 20) | PSNR: 29.25 -> 29.92 | Improved |
| use a single quantization-guided model instead of IDCN's quality-20 model on quality-10 images | Quantization Guided JPEG Artifact Correction · Live-1 generalization (image quality 10) | PSNR: 26.95 -> 27.65 | Improved |
| apply IDCN quality-10 model to quality-50 images versus leaving JPEG untouched | IDCN · Live-1 generalization (image quality 50) | PSNR: 30.91 -> 30.19 | Hurt |
| replace CFM layers with channelwise concatenation of the quantization matrix or ignoring it (ablation) | Quantization Guided JPEG Artifact Correction · Classic-5 quality 10 (Y channel) | PSNR | Improved |
| add BlockNet before FrequencyNet instead of FrequencyNet alone (ablation) | Quantization Guided JPEG Artifact Correction · Classic-5 quality 10 (Y channel) | PSNR | Improved |
| add the fusion layer (ablation) | Quantization Guided JPEG Artifact Correction · Classic-5 quality 10 (Y channel) | PSNR | Improved |
| add GAN loss training (Ours-GAN) to the regression model | Quantization Guided JPEG Artifact Correction · Live-1, BSDS500, ICB | FID | Improved |
Closest papers by meaning
- JPEG Artifacts Removal via Contrastive Representation Learning (ECCV 2022 · similarity 0.88)
- Task-Aware Quantization Network for JPEG Image Compression (ECCV 2020 · similarity 0.86)
- JPEG Inspired Deep Learning (ICLR 2025 · similarity 0.85)
- JDEC: JPEG Decoding via Enhanced Continuous Cosine Coefficients (CVPR 2024 · similarity 0.83)
- Frequency-Biased Synergistic Design for Image Compression and Compensation (CVPR 2025 · similarity 0.83)
- Learning from Loss Landscape: Generalizable Mixed-Precision Quantization via Adaptive Sharpness-Aware Gradient Aligning (ICML 2025 · similarity 0.82)
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.
