Research map / Image generation / Image restoration
A Trilateral Weighted Sparse Coding Scheme for Real-World Image Denoising
Jun Xu, Lei Zhang, David 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 weight matrix W2 (per-patch noise weights) to the data-fidelity term (TWSC vs WSC), AWGN sigma=15 (ablation) | WSC · AWGN sigma=15, 20 grayscale images | PSNR (dB): 32.06 -> 32.34 | Improved |
| add weight matrix W2 (TWSC vs WSC), AWGN sigma=75 (ablation) | WSC · AWGN sigma=75, 20 grayscale images | PSNR (dB): 24.54 -> 25.15 | Improved |
| add weight matrix W2 (TWSC vs WSC) (ablation) | WSC · Real-world noisy images, 15 cropped (Dataset 2) | PSNR (dB): 37.36 -> 37.81 | Improved |
| add weight matrix W2 (TWSC vs WSC) (ablation) | WSC · Real-world noisy images, 15 cropped (Dataset 2) | SSIM: 0.9516 -> 0.9586 | Improved |
| add weight matrix W2 (TWSC vs WSC) (ablation) | WSC · DND benchmark, 1000 cropped real-world images | PSNR (dB): 36.81 -> 37.94 | Improved |
| add weight matrix W2 (TWSC vs WSC) (ablation) | WSC · DND benchmark, 1000 cropped real-world images | SSIM: 0.9165 -> 0.9403 | Improved |
| use TWSC instead of MCWNNM | real-world denoising · DND benchmark, 1000 cropped real-world images | PSNR (dB): 37.38 -> 37.94 | Improved |
| use TWSC instead of MCWNNM | real-world denoising · Real-world noisy images, 15 cropped (Dataset 2) | PSNR (dB): 37.70 -> 37.81 | Improved |
| use TWSC instead of discriminatively trained DnCNN on AWGN | AWGN denoising · AWGN sigma=15, 20 grayscale images | PSNR (dB): 32.59 -> 32.34 | No clear effect |
| use TWSC instead of DnCNN on real noise | DnCNN · DND benchmark, 1000 cropped real-world images | PSNR (dB): 32.41 -> 37.94 | Improved |
| use TWSC instead of Neat Image (speed) | Neat Image · DND, time per 512x512 image | runtime (s): 1.1 -> 195.2 | Hurt |
| add weight matrix W2 (TWSC vs WSC) runtime cost (ablation) | WSC · DND, time per 512x512 image | runtime (s): 188.6 -> 195.2 | No clear effect |
Closest papers by meaning
- IDR: Self-Supervised Image Denoising via Iterative Data Refinement (CVPR 2022 · similarity 0.87)
- YOND: Practical Blind Raw Image Denoising Free from Camera-Specific Data Dependency (TPAMI 2026 · similarity 0.87)
- Convexity-Aware Noise Calibration: A Self-Supervised Framework for Noise-Level-Unknown Image Denoising (CVPR 2026 · similarity 0.86)
- Optimizing Image Compression via Joint Learning with Denoising (ECCV 2022 · similarity 0.86)
- Toward Convolutional Blind Denoising of Real Photographs (CVPR 2019 · similarity 0.86)
- Zero-Shot Blind-spot Image Denoising via Implicit Neural Sampling (CVPR 2025 · 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.
