Research map / Architectures and efficiency / Attention and transformers
Learned Image Compression with Hierarchical Progressive Context Modeling
Yuqi Li, Haotian Zhang, Li Li, Dong Liu
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 hierarchical context extraction (ablation) | HPCM-Base · Kodak | BD-Rate / kMACs per pixel: 1.07% -> 0.00%; 1107.48 -> 918.57 kMACs/pixel | Improved |
| add progressive context fusion (ablation) | HPCM-Base · Kodak | BD-Rate: 4.71% -> 0.00% | Improved |
| use cross-attention with C_i instead of entropy parameter state psi_i as progressive context (ablation) | HPCM-Base · Kodak | BD-Rate: 1.17% -> 0.00% | Improved |
| increase coding steps at largest scale from 3 to 6 (ablation) | HPCM-Base · Kodak | BD-Rate: 2.39% -> 0.00%; 663.90 -> 918.57 kMACs/pixel | Improved |
| increase coding steps at largest scale from 6 to 12 (ablation) | HPCM-Base · Kodak | BD-Rate: 0.00% -> -2.55%; 918.57 -> 1427.91 kMACs/pixel | Improved |
| use 2 coding steps instead of 4 at smallest scale (ablation) | HPCM-Base · Kodak | BD-Rate: 0.35% -> 0.00% | No clear effect |
| add third hierarchical stage over two stages (ablation) | HPCM-Base · Kodak | BD-Rate: 0.35% -> 0.00% | Improved |
| add fourth hierarchical stage (ablation) | HPCM-Base · Kodak | BD-Rate: 0.00% -> 0.18% | No clear effect |
| use non-shared parameters across scales (ablation) | HPCM-Base · Kodak | params / BD-Rate: 68.50 -> 189.99 M params; BD-Rate -0.15% | No clear effect |
| use HPCM entropy model instead of CHARM with same transform | HPCM-Base · Kodak | PSNR BD-Rate: 0.86% -> -15.31% | Improved |
| use HPCM entropy model instead of DCVC-DC intra with same transform | HPCM-Base · Kodak | PSNR BD-Rate: -9.18% -> -15.31% | Improved |
| use HPCM-Large instead of HPCM-Base (deeper transform and entropy networks) | HPCM · Tecnick | PSNR BD-Rate: -18.16% -> -22.20% | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- ELIC — ELIC: Efficient Learned Image Compression With Unevenly Grouped Space-Channel Contextual Adaptive Coding (CVPR 2022) · better on 2 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Linear Attention Modeling for Learned Image Compression (CVPR 2025 · similarity 0.87)
- MambaIC: State Space Models for High-Performance Learned Image Compression (CVPR 2025 · similarity 0.87)
- Diversify, Contextualize, and Adapt: Efficient Entropy Modeling for Neural Image Codec (NeurIPS 2024 · 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)
- Fix the Structural Bottleneck: Context Compression via Explicit Information Transmission (NeurIPS 2026 · similarity 0.86)
- GLIC: General Format Learned Image Compression (AAAI 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.
