Research map / Architectures and efficiency / RNNs and normalization
D3: Deep Dual-Domain Based Fast Restoration of JPEG-Compressed Images
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 the dual-domain, JPEG-prior D^3 architecture instead of AR-CNN | JPEG artifact restoration · LIVE1, Q = 5 | PSNR (dB): 26.64 -> 27.37 | Helped |
| Use the dual-domain, JPEG-prior D^3 architecture instead of AR-CNN | JPEG artifact restoration · LIVE1, Q = 10 | PSNR (dB): 29.03 -> 29.96 | Helped |
| Use the dual-domain, JPEG-prior D^3 architecture instead of AR-CNN | JPEG artifact restoration · LIVE1, Q = 20 | PSNR (dB): 31.30 -> 32.21 | Helped |
| Use D^3-256 instead of AR-CNN | JPEG artifact restoration · LIVE1, Q = 5 | running time (ms): 396.76 -> 12.20 | Helped |
| Increase dictionary dimensions from p_Φ = p_Ψ = 128 to p_Φ = p_Ψ = 256 (ablation) | D^3 · LIVE1, Q = 5 | PSNR (dB): 26.26 -> 27.37 | Helped |
| Increase dictionary dimensions from p_Φ = p_Ψ = 128 to p_Φ = p_Ψ = 256 (ablation) | D^3 · LIVE1, Q = 10 | PSNR (dB): 28.62 -> 29.96 | Helped |
| Increase dictionary dimensions from p_Φ = p_Ψ = 128 to p_Φ = p_Ψ = 256 (ablation) | D^3 · LIVE1, Q = 20 | PSNR (dB): 31.20 -> 32.21 | Helped |
| Replace D^3's task-specific design with a pixel-domain fully connected network using ReLU and dropout (ablation) | D^3-256 · LIVE1, Q = 5 | PSNR (dB): 25.83 vs 27.37 | Hurt |
| Replace D^3's task-specific design with a pixel-domain fully connected network using ReLU and dropout (ablation) | D^3-256 · LIVE1, Q = 10 | PSNR (dB): 28.24 vs 29.96 | Hurt |
| Replace D^3's task-specific design with a pixel-domain fully connected network using ReLU and dropout (ablation) | D^3-256 · LIVE1, Q = 20 | PSNR (dB): 31.27 vs 32.21 | Hurt |
| Use sparse-coding-based initialization instead of random initialization (ablation) | D^3 · JPEG artifact restoration, Q = 5 visual examples | detail preservation | Helped |
| Add the signal-dependent box-constrained loss on DCT coefficients (ablation) | D^3 · JPEG artifact restoration, Q = 5 visual examples | PSNR | Helped |
Closest papers by meaning
- DDR: Exploiting Deep Degradation Response as Flexible Image Descriptor (NeurIPS 2024 · similarity 0.85)
- Improving the Learning Capability of Small-size Image Restoration Network by Deep Fourier Shifting (NeurIPS 2024 · similarity 0.85)
- ResDiff: Combining CNN and Diffusion Model for Image Super-resolution (AAAI 2024 · similarity 0.85)
- Dual Prior Unfolding for Snapshot Compressive Imaging (CVPR 2024 · similarity 0.84)
- C3: High-Performance and Low-Complexity Neural Compression from a Single Image or Video (CVPR 2024 · similarity 0.84)
- DCTMamba: Advancing JPEG Image Restoration Through Long-Sequence Modeling and Adaptive Frequency Strategy (AAAI 2025 · similarity 0.84)
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.