Research map / Image generation / Image restoration
Learning Content-Weighted Deep Image Compression
Mu Li, Wangmeng Zuo, Shuhang Gu, Jane You, David D. Zhang
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 channel-wise learnt quantization (LCMQ) instead of one shared learnt quantizer (LMQ) (ablation) | CWIC · Kodak | quantization error (average): 1.05 -> 0.90 | Improved |
| add learnt quantization (LMQ) instead of fixed uniform levels (FMQ) (ablation) | CWIC · Kodak | quantization error (average): 1.97 -> 1.05 | Improved |
| use 8-level multi-valued quantization (LCMQ) instead of binarization (BIN) (ablation) | CWIC · Kodak | quantization error (average): 3.18 -> 0.90 | Improved |
| use LCMQ instead of BIN, LMQ or FMQ (ablation) | CWIC · Kodak | MS-SSIM rate-distortion | Improved |
| add importance map subnet with spatially variant bit allocation (ablation) | CWIC · Kodak | MS-SSIM rate-distortion | Improved |
| use TCAE or inclined TCAE instead of modified CABAC context (ablation) | CWIC · Kodak | lossless compression ratio | Improved |
| use larger context (TCAE 37x37) instead of 4x5x5 context of Li et al. (ablation) | CWIC · Kodak | lossless compression ratio | Improved |
| increase number of feature-map groups in inclined TCAE (1,2,4,8 for o'; 4,8,16,32 for p') (ablation) | inclined TCAE · Kodak | compression ratio | Improved |
| use inclined TCAE instead of TCAE for decoding o' (ablation) | CWIC · Kodak | decoding time (s): 202.8 -> 0.923 | Improved |
| use inclined TCAE instead of TCAE for decoding p' (ablation) | CWIC · Kodak | decoding time (s): 6.58 -> 0.161 | Improved |
| train with MS-SSIM loss instead of MSE loss | CWIC · Kodak | MS-SSIM rate-distortion | Improved |
Closest papers by meaning
- Learning Convolutional Networks for Content-Weighted Image Compression (CVPR 2018 · similarity 0.94)
- Entroformer: A Transformer-based Entropy Model for Learned Image Compression (ICLR 2022 · similarity 0.87)
- Improved Lossy Image Compression With Priming and Spatially Adaptive Bit Rates for Recurrent Networks (CVPR 2018 · similarity 0.87)
- Conditional Entropy Coding for Efficient Video Compression (ECCV 2020 · similarity 0.86)
- Conditional Probability Models for Deep Image Compression (CVPR 2018 · similarity 0.86)
- Laplacian-guided Entropy Model in Neural Codec with Blur-dissipated Synthesis (CVPR 2024 · similarity 0.85)
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.
