Research map / Image generation / Image restoration
Deep Lossy Plus Residual Coding for Lossless and Near-Lossless Image Compression
Yuanchao Bai, Xianming Liu, Kai Wang, Xiangyang Ji, Xiaolin Wu, Wen Gao
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 proposed analysis/synthesis blocks instead of 5x5 conv in LIC (ablation) | DLPR · ImageNet64 (lossless) | bpsp: 3.75 -> 3.71 | Improved |
| add Swin attention blocks instead of standard attention (ablation) | DLPR · ImageNet64 (lossless) | bpsp: 3.71 -> 3.69 | Improved |
| add Swin attention and A/S blocks versus 5x5 conv (ablation) | DLPR · DIV2K (lossless) | bpsp: 2.60 -> 2.55 | Improved |
| add context C_r to residual compressor (ablation) | DLPR · ImageNet64 (lossless) | bpsp: 4.03 -> 3.69 | Improved |
| add feature u to residual compressor (ablation) | DLPR · ImageNet64 (lossless) | bpsp: 3.78 -> 3.69 | Improved |
| use logistic mixture instead of Gaussian single model (ablation) | DLPR · ImageNet64 (lossless) | bpsp: 3.78 -> 3.69 | Improved |
| use logistic mixture instead of Gaussian mixture (ablation) | DLPR · ImageNet64 (lossless) | bpsp: 3.70 -> 3.69 | No clear effect |
| increase lambda from 0 to 0.06 (ablation) | DLPR · Kodak (lossless) | total rate: 2.86 -> 3.02 | Hurt |
| train scalable model instead of tau-specific models with straight-through (ablation) | DLPR · near-lossless Kodak | bpsp | Improved |
| add SQRC bias correction (ablation) | DLPR · near-lossless Kodak | bit rate | Improved |
| switch from lossless to tau=1 near-lossless mode (ablation) | DLPR · Kodak | bit rate: about 39% saved | Improved |
| use lambda=0.03 instead of lambda=0 for near-lossless (ablation) | DLPR · near-lossless | bit rate | Improved |
Closest papers by meaning
- Learning Scalable lY=-Constrained Near-Lossless Image Compression via Joint Lossy Image and Residual Compression (CVPR 2021 · similarity 0.92)
- Learning Better Lossless Compression Using Lossy Compression (CVPR 2020 · similarity 0.90)
- Practical Full Resolution Learned Lossless Image Compression (CVPR 2019 · similarity 0.88)
- Towards Efficient Image Compression Without Autoregressive Models (NeurIPS 2023 · similarity 0.88)
- Ultra Lowrate Image Compression with Semantic Residual Coding and Compression-aware Diffusion (ICML 2025 · similarity 0.86)
- Large Language Models for Lossless Image Compression: Next-Pixel Prediction in Language Space is All You Need (NeurIPS 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.
