Research map / Image generation / Image restoration
Turbo-DDCM: Fast and Flexible Zero-Shot Diffusion-Based Image Compression
Amit Vaisman, Guy Ohayon, Hila Manor, Michael Elad, Tomer Michaeli
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 |
|---|---|---|---|
| combine many codebook atoms per step via closed-form thresholding instead of matching pursuit | DDCM · Kodak24 round-trip time | seconds: 65 -> 1.8 | Improved |
| combine many atoms per step (Turbo-DDCM) | DDCM · number of diffusion steps | reduction: more than 92% fewer steps | Improved |
| use combinatorial index bit protocol instead of ordered index encoding | Turbo-DDCM · BPP | BPP: ~40% lower BPP | Improved |
| allow negative coefficients (sign selection) vs convex combination | Turbo-DDCM · representational capacity | residual approximation | Improved |
| Turbo-DDCM vs DDCM at ~0.05 bpp | DDCM · Kodak24 | PSNR: 22.84 (0.046 bpp) -> 23.82 (0.058 bpp) | Improved |
| Turbo-DDCM vs PSC | PSC · Kodak24 | FID: 40.70 -> 25.75 | Improved |
| Turbo-DDCM vs DiffC at low bitrate | DiffC · Kodak24 | FID: 26.39 (DiffC 0.057) vs 25.75 (Turbo 0.058) | Mixed |
| Turbo-DDCM vs DiffC runtime | DiffC · Kodak24 | round-trip time (s): 11.23 -> 1.97 | Improved |
| increase bitrate to high regime due to SD 2.1 latent bound | Turbo-DDCM · Kodak24 high bitrate | PSNR | Hurt |
Closest papers by meaning
- Compressed Image Generation with Denoising Diffusion Codebook Models (ICML 2025 · similarity 0.91)
- Lossy Compression with Pretrained Diffusion Models (ICLR 2025 · similarity 0.89)
- Zero-shot Denoising via Neural Compression: Theoretical and algorithmic framework (NeurIPS 2025 · similarity 0.89)
- OSCAR: One-Step Diffusion Codec Across Multiple Bit-rates (NeurIPS 2025 · similarity 0.88)
- Diffusion Models Without Attention (CVPR 2024 · similarity 0.87)
- Masked Pre-training Enables Universal Zero-shot Denoiser (NeurIPS 2024 · similarity 0.87)
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.
