Research map / Image generation / Image restoration
How To Exploit the Transferability of Learned Image Compression to Conventional Codecs
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 |
|---|---|---|---|
| Add surrogate-guided pre-filtering optimized for MS-SSIM | Conventional codecs · Kodak | rate savings: over 20% | Helped |
| Add surrogate-guided pre-filtering optimized for MS-SSIM | Conventional codecs · CLIC Professional validation | rate savings: up to 15% | Helped |
| Use classifier feedback to train a pre-filter for JPEG | ImageNet image recognition compression · ImageNet | classification performance | Helped |
| Add GAN-guided filtering to H265 input | H265 · Visual comparison examples | rate (bit/pixel): 0.086 bit/pixel -> 0.084 bit/pixel; 0.045 bit/pixel -> 0.042 bit/pixel; 0.190 bit/pixel -> 0.192 bit/pixel | Mixed |
| Use an MS-SSIM-trained surrogate instead of an MSE-trained surrogate (ablation) | Surrogate-induced filtering · Kodak | rate savings | No clear effect |
| Initialize the task-aware filter with high rather than very low variance (ablation) | Task-aware surrogate-induced filtering · ImageNet image recognition | classification performance | Hurt |
| Use the surrogate reconstruction rather than the target codec reconstruction for classifier feedback (ablation) | Task-aware surrogate-induced filtering · ImageNet image recognition | classification performance | Hurt |
| Use a GAN loss weight low enough to allow only small image changes (ablation) | GAN-guided filtering · Low-rate image compression | texture replacement | Hurt |
Closest papers by meaning
- High-Fidelity Generative Image Compression (NeurIPS 2020 · similarity 0.92)
- End-to-end Optimized Image Compression (ICLR 2017 · similarity 0.90)
- Real-Time Adaptive Image Compression (ICML 2017 · similarity 0.89)
- GLIC: General Format Learned Image Compression (AAAI 2025 · similarity 0.88)
- Controllable Distortion-Perception Tradeoff Through Latent Diffusion for Neural Image Compression (AAAI 2025 · similarity 0.88)
- Good, Cheap, and Fast: Overfitted Image Compression with Wasserstein Distortion (CVPR 2025 · 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.