Research map / Image generation / Image restoration
denoiSplit: a method for joint microscopy image splitting and unsupervised denoising
Ashesh Ashesh, Florian Jug
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 |
|---|---|---|---|
| weight KL loss by summing over pixels rather than averaging per hierarchy level (ablation) | muSplit-based VSE network · BioSR and other splitting tasks | PSNR / MS-SSIM: 28.7 -> 34.7 PSNR (T8, sigma 1, muSplit vs uSplit KL) | Improved |
| add pixel noise models to the likelihood for joint denoising (ablation) | uSplit with changed KL · Actin vs Mito high-SNR (sigma 1) | PSNR (dB): 34.7 -> 35.5 | Improved |
| add pixel noise models to the likelihood for joint denoising (ablation) | uSplit with changed KL · Actin vs Mito high-SNR (sigma 4) | PSNR (dB): 32.3 -> 32.2 | No clear effect |
| joint denoiSplit versus sequential HDN then muSplit | Actin vs Tubulin (T7) · PaviaATN Actin vs Tubulin sigma 1 | PSNR (dB): 27.8 -> 26.4 | Hurt |
| replace muSplit with denoiSplit on noisy data | muSplit · PaviaATN Actin vs Tubulin sigma 4 | PSNR (dB): 19.1 -> 25.2 | Improved |
| compute noise model with different procedures (microscope access, clean-noisy pair, N2V denoised) (ablation) | denoiSplit · ER vs CCPs and ER vs MT | PSNR | No clear effect |
| increase number of posterior samples for uncertainty estimate (ablation) | denoiSplit calibration · calibration plot | calibration closeness to y=x | Improved |
| joint splitting versus unsupervised denoising only (HDN) | denoiSplit · BioSR tasks | PSNR | Hurt |
Closest papers by meaning
- Fully Unsupervised Diversity Denoising with Convolutional Variational Autoencoders (ICLR 2021 · similarity 0.90)
- A Poisson-Gaussian Denoising Dataset With Real Fluorescence Microscopy Images (CVPR 2019 · similarity 0.90)
- Interpretable Unsupervised Diversity Denoising and Artefact Removal (ICLR 2022 · similarity 0.89)
- scSplit: Bringing Severity Cognizance to Image Decomposition in Fluorescence Microscopy (NeurIPS 2025 · similarity 0.88)
- Efficient Online Training for Zero-Shot Time-Lapse Microscopy Denoising and Super-Resolution (AAAI 2025 · similarity 0.87)
- Interpretable Diffusion via Information Decomposition (ICLR 2024 · 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.
