Research map / Architectures and efficiency / CNN design
Multiscale Adaptive Representation of Signals: I. The Basic Framework
Cheng Tai, Weinan E
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 |
|---|---|---|---|
| learn AdaFrame filters from clean images instead of the noisy image | AdaFrame denoising · Single image denoising (sigma=0.1) | PSNR (dB): 28.8 -> 29.3 | Improved |
| learn filters from clean images instead of noisy images | AdaFrame · Extended Yale B (noisy, sigma=0.1) | PSNR (dB): 31.35 -> 32.07 | Improved |
| learn filters from clean images instead of noisy images | K-SVD · Extended Yale B (noisy, sigma=0.1) | PSNR (dB): 31.4 -> 32.01 | Improved |
| replace K-SVD with AdaFrame (noisy-learned filters) | denoising · Single image denoising (sigma=0.1) | PSNR (dB): 28.65 -> 28.8 | Improved |
| increase AdaFrame filter support from 8x8 to 12x12 (ablation) | AdaFrame denoising · Barbara sigma=0.05 | PSNR (dB): 31.87 -> 33.22 | Improved |
| increase AdaFrame filter support from 8x8 to 12x12 (ablation) | AdaFrame denoising · House sigma=0.1 | PSNR (dB): 31.19 -> 31.45 | Improved |
| increase K-SVD patch size from 8x8 to 12x12 (ablation) | K-SVD denoising · Boat sigma=0.1 | PSNR (dB): 29.19 -> 28.70 | Hurt |
| use AdaFrame 12x12 instead of K-SVD 12x12 | denoising · Lena sigma=0.1 | PSNR (dB): 30.84 -> 31.39 | Improved |
| use AdaFrame 8x8 instead of K-SVD 8x8 | denoising · House sigma=0.1 | PSNR (dB): 32.15 -> 31.19 | Hurt |
| use AdaFrame instead of K-SVD for speed | denoising runtime · Denoising runtime (laptop, 8x8 support) | denoising time (s): 6.5s -> 0.6s | Improved |
| use AdaFrame filters instead of K-SVD dictionary training | training runtime · Denoising runtime (laptop, 8x8 support) | training time (s): 25s -> 3.7s | Improved |
| increase number of filters from 6 to 12 (ablation) | AdaFrame features · MNIST | precision: 97.0 % -> 97.4% | Improved |
Closest papers by meaning
- A Dynamic Frame Selection Framework for Fast Video Recognition (TPAMI 2020 · similarity 0.84)
- AdaFrame: Adaptive Frame Selection for Fast Video Recognition (CVPR 2019 · similarity 0.84)
- AdaCoF: Adaptive Collaboration of Flows for Video Frame Interpolation (CVPR 2020 · similarity 0.83)
- Efficient Multi-scale Network with Learnable Discrete Wavelet Transform for Blind Motion Deblurring (CVPR 2024 · similarity 0.82)
- WST: Wavelet-Based Multi-scale Tuning for Visual Transfer Learning (AAAI 2025 · similarity 0.81)
- Diversify, Contextualize, and Adapt: Efficient Entropy Modeling for Neural Image Codec (NeurIPS 2024 · similarity 0.81)
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.
