Research map / Image generation / Image restoration
Block-based Learned Image Compression without Blocking Artifacts
Jong Wook Kim, Suyong Bahk, TaeHwa Lee, HyunDong Cho, Donghyun Kim, Sung-Chang Lim, Jin Soo Choi, Hui Yong Kim
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 |
|---|---|---|---|
| reduce minimal overlap by one pixel in hyper-decoder hs (ablation) | Hyperprior w/o AR block-based LIC · 2K, 256x256 blocks | BD-rate: +230.52% | Hurt |
| reduce minimal overlap by one pixel in ha (ablation) | Hyperprior w/o AR block-based LIC · 2K, 256x256 blocks | BD-rate: +76.34% | Hurt |
| reduce minimal overlap by one pixel in gs (ablation) | Hyperprior w/o AR block-based LIC · 2K, 256x256 blocks | BD-PSNR (dB): -20.31 dB | Hurt |
| reduce minimal overlap by one pixel in ga (ablation) | Hyperprior w/o AR block-based LIC · 2K, 256x256 blocks | BD-PSNR (dB): -14.36 dB | Hurt |
| use analytically derived minimal overlap (ablation) | four CNN LIC models · 2K and 4K images | BD-rate vs full image: -0.004% to +0.0081% | No clear effect |
| use minimal overlap and avoid extra zero padding instead of JPEG-AI empirical overlap | JPEG-AI BOP · 2K 192x192 | BD-rate (Y): 0.79 -> <=0.01 | Improved |
| use minimal overlap instead of JPEG-AI empirical overlap | JPEG-AI BOP · 4K 192x192 | peak encoder MAC/pel: 87.6 -> 70.7 | Improved |
| use minimal overlap instead of JPEG-AI empirical overlap | JPEG-AI BOP · 2K 192x192 | decoding time (s): 1.24 -> 1.45 | Hurt |
| block-based decoding with minimal overlap vs full-image coding | four CNN LIC models · 4K, 256x256 and 512x512 blocks | peak memory: to 13.94% (encoder), 13.33% (decoder) of full image | Improved |
Closest papers by meaning
- Learned Image Compression With Mixed Transformer-CNN Architectures (CVPR 2023 · similarity 0.86)
- JPEG Artifacts Reduction via Deep Convolutional Sparse Coding (ICCV 2019 · similarity 0.86)
- Learning Dual Priors for JPEG Compression Artifacts Removal (ICCV 2021 · similarity 0.86)
- Learning End-to-End Lossy Image Compression: A Benchmark (TPAMI 2021 · similarity 0.85)
- Balanced Rate-Distortion Optimization in Learned Image Compression (CVPR 2025 · similarity 0.84)
- Block Image Compressive Sensing with Local and Global Information Interaction (AAAI 2024 · 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.
