Research map / Image generation / Autoregressive and latent generation
Learning Asymmetric and Local Features in Multi-Dimensional Data Through Wavelets With Recursive Partitioning
Meng Li, Li Ma
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 hierarchical adaptive partition to basic wavelet regression (1D-Haar to WARP) | wavelet regression · ImageNet 512x512 (100 images) | MSE and MAE | Improved |
| use WARP instead of pre-trained DnCNN at sigma=0.4 | denoising · BSD68 (sigma=0.4) | MSE (x1e-3): 14.90 -> 6.10 | Improved |
| use WARP instead of pre-trained DnCNN at sigma=0.6 | denoising · BSD68 (sigma=0.6) | MSE (x1e-3): 71.65 -> 7.88 | Improved |
| use WARP instead of pre-trained DnCNN at sigma=0.2 | denoising · BSD68 (sigma=0.2) | MSE (x1e-3): 3.35 -> 3.75 | Hurt |
| use WARP instead of 2D DWT for energy concentration | wavelet representation · ImageNet images | number of coefficients: 80%, 70%, 70% fewer coefficients | Improved |
| use WARP instead of 1D DWT for energy concentration | wavelet representation · ImageNet images | number of coefficients: 97%, 99%, 90% fewer coefficients | Improved |
| use WARP instead of SSR on foveal OCT images | retinal denoising · Retinal OCT, 18 foveal images | mean PSNR: 28.10 -> 28.18 | Improved |
| use full separate hyperparameters (WARP-full) instead of tuned WARP (ablation) | WARP · Simulation n=64, f2, sigma=0.2 | average MSE (x1e-2): 0.12 -> 0.11 | No clear effect |
| vary tau and eta prior parameterization (ablation) | WARP · Shepp-Logan 256x256, sigma=0.7 | average MSE (x1e-2): 0.89 -> 0.94 | No clear effect |
| increase image size from 1024x1024 to 4096x4096 | WARP · Scalability | running time: 5.3 seconds -> about 2 minutes | No clear effect |
Closest papers by meaning
- Likelihood Training of Cascaded Diffusion Models via Hierarchical Volume-preserving Maps (ICLR 2024 · similarity 0.84)
- Wavelet-Driven Masked Image Modeling: A Path to Efficient Visual Representation (AAAI 2025 · similarity 0.84)
- Provably Robust Score-Based Diffusion Posterior Sampling for Plug-and-Play Image Reconstruction (NeurIPS 2024 · similarity 0.84)
- A Unified Degradation-Robust Approach to SSL and UDA for 3D Medical Images (AAAI 2025 · similarity 0.84)
- Asymmetric Flow Models (NeurIPS 2026 · similarity 0.84)
- Task-Driven Wavelets using Constrained Empirical Risk Minimization (CVPR 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.
