Research map / Architectures and efficiency / RNNs and normalization
JPEG Processing Neural Operator for Backward-Compatible Coding
Woo Kyoung Han, Yongjun Lee, Byeonghun Lee, Sang Hyun Park, Sunghoon Im, Kyong Hwan Jin
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 JDNO neural decoder (JPNeO) in place of the JPEG decoder at low bpp (ablation) | JPEG pipeline · Quantization matrix comparison (low bpp) | PSNR (dB): 25.29 -> 27.77 | Improved |
| add JENO neural encoder in front of the JPEG encoder at high bpp (ablation) | JPEG pipeline · Quantization matrix comparison (high bpp) | PSNR (dB): 33.07 -> 33.19 | Improved |
| add both JENO and JDNO at high bpp (ablation) | JPEG pipeline · Quantization matrix comparison (high bpp) | PSNR (dB): 33.07 -> 35.95 | Improved |
| use learned quantization matrix Q_psi instead of standard Q with plain JPEG (ablation) | JPEG pipeline · Quantization matrix comparison (high bpp) | PSNR (dB): 33.07 -> 34.03 | Improved |
| use learned quantization matrix Q_psi instead of Q with full JPNeO (ablation) | JPNeO · Quantization matrix comparison (high bpp) | PSNR (dB): 35.95 -> 36.86 | Improved |
| use learned quantization matrix Q_psi instead of Q with full JPNeO at low bpp (ablation) | JPNeO · Quantization matrix comparison (low bpp) | PSNR (dB): 27.77 -> 27.77 | No clear effect |
| use JDNO decoder instead of CNN decoder (ablation) | JPNeO · LIVE-1 (q=0) | PSNR (dB): 23.14 -> 23.25 | Improved |
| use JDNO decoder instead of U-Net decoder (ablation) | JPNeO · LIVE-1 (q=0) | PSNR (dB): 22.77 -> 23.25 | Improved |
| use JENO encoder instead of CNN encoder (ablation) | JPNeO · LIVE-1 (q=90) | PSNR (dB): 37.17 -> 37.20 | No clear effect |
| use JENO encoder instead of U-Net encoder (ablation) | JPNeO · LIVE-1 (q=100) | PSNR (dB): 44.87 -> 45.47 | Improved |
| use the full JPNeO instead of the lightweight JPNeO- (ablation) | JPNeO · LIVE-1 (q=0) | PSNR (dB): 22.98 -> 23.15 | Improved |
| use JPNeO instead of FBCNN at lowest quality | FBCNN · LIVE-1 (q=0) | PSNR (dB): 21.70 -> 23.15 | Improved |
Closest papers by meaning
- Computationally-Efficient Neural Image Compression with Shallow Decoders (ICCV 2023 · similarity 0.86)
- JDEC: JPEG Decoding via Enhanced Continuous Cosine Coefficients (CVPR 2024 · similarity 0.85)
- Neural Data-Dependent Transform for Learned Image Compression (CVPR 2022 · similarity 0.84)
- Full Resolution Image Compression With Recurrent Neural Networks (CVPR 2017 · similarity 0.84)
- JPEG Artifacts Removal via Contrastive Representation Learning (ECCV 2022 · similarity 0.84)
- JPEG Inspired Deep Learning (ICLR 2025 · similarity 0.82)
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.
