Research map / Image generation / Image restoration
Image Demoireing with Learnable Bandpass Filters
Bolun Zheng, Shanxin Yuan, Gregory Slabaugh, Ales Leonardis
Check 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 |
|---|---|---|---|
| Use MBCNN rather than CAS-CNN | Image demoireing models · LCDMoire validation | PSNR: +7.88dB | Improved |
| Use MBCNN rather than the second-best method | Image demoireing models · TIP2018 | PSNR: +2.28 dB | Improved |
| Add the DCT-domain transform while retaining fixed passbands (ablation) | MBCNN-nDDT · LCDMoire validation | PSNR: 0.18dB | Improved |
| Add learnable passband weights (ablation) | MBCNN-nLP · LCDMoire validation | PSNR: 0.95 dB | Improved |
| Use block-IDCT size p=8 instead of p=6, p=10, or p=12 (ablation) | MBCNN · LCDMoire validation | PSNR: 43.25 & 44.04 & 43.45 & 43.17 | Mixed |
| Add classic Sobel loss to L1 (ablation) | MBCNN with L1 loss · LCDMoire validation | PSNR: 41.83 -> 43.64 | Improved |
| Add the two directional filters of ASL to the Sobel-loss setup (ablation) | MBCNN with L1 + Sobel · LCDMoire validation | PSNR: 0.40dB | Improved |
| Replace ASL with Laplace loss (ablation) | MBCNN with L1 + ASL · LCDMoire validation | PSNR: 44.04 -> 42.92 | Hurt |
| Replace ASL with SSIM loss (ablation) | MBCNN with L1 + ASL · LCDMoire validation | PSNR: 44.04 -> 43.36 | Hurt |
| Replace ASL with perceptual loss (ablation) | MBCNN with L1 + ASL · LCDMoire validation | PSNR: 44.04 -> 43.83 | Hurt |
| Reduce n_G to 64 and n_D to 32 (ablation) | MBCNN · LCDMoire validation | PSNR: -1.46 dB | Hurt |
| Apply geometric self-ensemble using rotated inputs | MBCNN · LCDMoire validation | PSNR: 44.04 -> 33.65 | Hurt |
Closest papers by meaning
- Learning Frequency Domain Priors for Image Demoireing (TPAMI 2021 · similarity 0.95)
- Wavelet-Based Dual-Branch Network for Image Demoiréing (ECCV 2020 · similarity 0.86)
- Self-Adaptively Learning to Demoiré from Focused and Defocused Image Pairs (NeurIPS 2020 · similarity 0.86)
- Freqformer: Image-Demoiréing Transformer via Effective Frequency Decomposition (ECCV 2026 · similarity 0.85)
- Efficient RAW Image Deblurring with Adaptive Frequency Modulation (NeurIPS 2025 · similarity 0.83)
- Direction-Aware Video Demoiréing with Temporal-Guided Bilateral Learning (AAAI 2024 · similarity 0.83)
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.
