Research map / Image generation / Image restoration
NGPS: Structure-Preserving Self-Supervised Denoising via Neighbor-Guided Patch Sampling
Jaehyun Cho, YOUNGJOON YOO
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 regional consistency loss (ablation) | NGPS · AAPM-Mayo | PSNR (dB): 36.50 -> 36.68 | Improved |
| add inter-slice continuity loss to NGPS (ablation) | NGPS · AAPM-Mayo | PSNR (dB): 36.68 -> 36.58 | Hurt |
| add regional consistency and IC losses (ablation) | NS-N2N · AAPM-Mayo | PSNR (dB): 35.43 -> 35.91 | Improved |
| use filtered guide for patch matching instead of raw noisy image (ablation) | NGPS · LIDC-IDRI | PSNR (dB): 28.94 -> 31.03 | Improved |
| use bilateral plus median guide instead of NLM guide (ablation) | NGPS · LIDC-IDRI | guide time (s per triplet): 12.6936 -> 0.1422 | Improved |
| retrieve from adjacent slice with guide-based matching instead of same-slice PixelBank (ablation) | NGPS · AAPM-Mayo | PSNR (dB): 35.16 -> 36.68 | Improved |
| match on filtered guide instead of adjacent-slice raw-patch bank (ablation) | NGPS · LIDC-IDRI 2.5 mm | PSNR (dB): 30.18 -> 30.92 | Improved |
| increase Top-K ensemble from 1 or 2 to K=4 (ablation) | NGPS · LIDC-IDRI | PSNR (dB): best at K=4 (PSNR 31.03 dB, SSIM 0.8102) | Improved |
| increase Top-K to 8 or 16 (ablation) | NGPS · LIDC-IDRI | PSNR (dB): K=4 31.03 dB best | Hurt |
| increase search window from 15 to 23 (ablation) | NGPS · LIDC-IDRI 1.25 mm | PSNR (dB): 31.08 -> 30.93 | Hurt |
| thicker slices (1.25 mm to 2.5 mm) | NGPS vs NS-N2N · LIDC-IDRI | PSNR drop (dB): -0.16 dB vs -0.63 dB | Improved |
| increase through-plane gap to 5 mm | NGPS vs NS-N2N · AAPM-Mayo | PSNR (dB): 34.48 vs 35.12 | Hurt |
Closest papers by meaning
- Score-based Self-supervised MRI Denoising (ICLR 2025 · similarity 0.87)
- Patch2Self2: Self-supervised Denoising on Coresets via Matrix Sketching (CVPR 2024 · similarity 0.86)
- Self-Supervised Diffusion MRI Denoising via Iterative and Stable Refinement (ICLR 2025 · similarity 0.85)
- Noise-Consistent Siamese-Diffusion for Medical Image Synthesis and Segmentation (CVPR 2025 · similarity 0.84)
- Zero-shot Denoising via Neural Compression: Theoretical and algorithmic framework (NeurIPS 2025 · similarity 0.84)
- A Unified Model for Compressed Sensing MRI Across Undersampling Patterns (CVPR 2025 · 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.
