Research map / Image generation / Fast diffusion sampling
Compression-Aware One-Step Diffusion Model for JPEG Artifact Removal
Jinpei Guo, Zheng Chen, Wenbo Li, Yong Guo, Yulun Zhang
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 CaVE embeddings as prompt instead of empty string (ablation) | CODiff · LIVE-1 | MUSIQ: 62.56 -> 67.13 | Improved |
| use CaVE instead of DAPE for prompt embedding (ablation) | CODiff · LIVE-1 | M-IQA: 0.3793 -> 0.4584 | Improved |
| use learnable embeddings instead of empty prompt (ablation) | CODiff · LIVE-1 | MUSIQ: 62.56 -> 63.39 | No clear effect |
| use CaVE instead of empty prompt (ablation) | CODiff · DIV2K-Val | MUSIQ: 56.68 -> 61.83 | Improved |
| add implicit reconstruction learning to explicit QF learning in CaVE (ablation) | CaVE · LIVE-1 | MUSIQ: 62.39 -> 67.13 | Improved |
| add explicit QF learning to implicit-only CaVE (ablation) | CaVE · LIVE-1 | MUSIQ: 64.74 -> 67.13 | Improved |
| add DISTS loss to L2 loss (ablation) | CODiff · LIVE-1/DIV2K ablation | LPIPS: 0.3976 -> 0.2109 | Improved |
| add GAN loss to L2 + DISTS (ablation) | CODiff · LIVE-1/DIV2K ablation | MUSIQ: 72.69 -> 73.34 | Improved |
| add GAN loss to L2 + DISTS (ablation) | CODiff · LIVE-1/DIV2K ablation | LPIPS: 0.2109 -> 0.2113 | No clear effect |
| add GAN loss to L2 only (ablation) | CODiff · LIVE-1/DIV2K ablation | LPIPS: 0.3976 -> 0.3899 | No clear effect |
| replace OSEDiff-style DAPE prompt extractor with lightweight CaVE | one-step diffusion · Complexity 1024x1024 | Params (G): 1.40 -> 1.00 | Improved |
| one-step CODiff instead of 50-step DiffBIR | diffusion JPEG restoration · Complexity 1024x1024 | inference time (s): 50.81 -> 0.57 | Improved |
Closest papers by meaning
- JPEG Artifacts Removal via Contrastive Representation Learning (ECCV 2022 · similarity 0.92)
- SODiff:Semantic-Oriented Diffusion Model for JPEG Compression Artifacts Removal (AAAI 2026 · similarity 0.91)
- Learning Dual Priors for JPEG Compression Artifacts Removal (ICCV 2021 · similarity 0.90)
- JPEG Artifacts Reduction via Deep Convolutional Sparse Coding (ICCV 2019 · similarity 0.89)
- OSCAR: One-Step Diffusion Codec Across Multiple Bit-rates (NeurIPS 2025 · similarity 0.88)
- Compressed Image Generation with Denoising Diffusion Codebook Models (ICML 2025 · 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.
