Research map / Image generation / Image restoration
Bitstream-Corrupted JPEG Images Are Restorable: Two-Stage Compensation and Alignment Framework for Image Restoration
Wenyang Liu, Yi Wang, Kim-Hui Yap, Lap-Pui Chau
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 SCA before GCA, including segment compensation and block alignment | JPEG restoration framework · Cityscapes 512×256 | PSNR: 12.41 -> 14.34 | Improved |
| Use GCA instead of EPDN after SCA | SCA restoration output · Cityscapes 512×256 | PSNR: 33.51 -> 42.05 | Improved |
| Use GCA instead of EPDN after SCA | SCA restoration output · Cityscapes 1024×512 | PSNR: 33.25 -> 40.09 | Improved |
| Use GCA instead of EPDN after SCA | SCA restoration output · Cityscapes 2048×1024 | PSNR: 33.40 -> 38.92 | Improved |
| Add SCA segment detection and normalization (ablation) | SCA + GCA · Cityscapes 512×256 | PSNR: 33.88 -> 42.05 | Improved |
| Add SCA segment detection and normalization (ablation) | SCA + GCA · Cityscapes 1024×512 | PSNR: 33.52 -> 40.09 | Improved |
| Add block alignment to SCA (ablation) | SCA + GCA · Cityscapes 512×256 | PSNR: 36.50 -> 42.05 | Improved |
| Add block alignment to SCA (ablation) | SCA + GCA · Cityscapes 1024×512 | PSNR: 35.01 -> 40.09 | Improved |
| Use Laplacian fusion instead of EPDN's pooling-enhanced module (ablation) | GCA · Cityscapes 512×256 | PSNR: 36.65 -> 42.05 | Improved |
| Use Laplacian fusion instead of EPDN's pooling-enhanced module (ablation) | GCA · Cityscapes 1024×512 | PSNR: 36.16 -> 40.09 | Improved |
| Add edge and Charbonnier losses instead of using original ℓ2 fidelity loss (ablation) | GCA training · Cityscapes 512×256 | PSNR: 41.27 -> 42.05 | Improved |
| Add edge and Charbonnier losses instead of using original ℓ2 fidelity loss (ablation) | GCA training · Cityscapes 1024×512 | PSNR: 39.53 -> 40.09 | Improved |
Closest papers by meaning
- OAPT: Offset-Aware Partition Transformer for Double JPEG Artifacts Removal (ECCV 2024 · similarity 0.85)
- Towards Flexible Blind JPEG Artifacts Removal (ICCV 2021 · similarity 0.85)
- D3: Deep Dual-Domain Based Fast Restoration of JPEG-Compressed Images (CVPR 2016 · similarity 0.85)
- DCTMamba: Advancing JPEG Image Restoration Through Long-Sequence Modeling and Adaptive Frequency Strategy (AAAI 2025 · similarity 0.83)
- Blind Bitstream-corrupted Video Recovery via Metadata-guided Diffusion Model (CVPR 2025 · similarity 0.83)
- JDEC: JPEG Decoding via Enhanced Continuous Cosine Coefficients (CVPR 2024 · 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.
