Scintillator Decorrelator

This repository contains code and Jupyter notebook examples for the publication:

Scintillator decorrelation for self-supervised X-ray radiograph denoising 📄 DOI: 10.1088/1361-6501/addc06

The article describes how to remove correlations from X-ray radiographs acquired by e.g. Caesium-Iodine scintillator detectors, via PRF estimation and deconvolution. This enables self-supervised radiograph denoising with Noise2Self and Noise2Void, as well as self-supervised Computed Tomography (CT) with a blind-spot losses in the sinogram.

📑 View slides.pdf — Supplementary presentation material.

---

📓 Notebooks

• 📘 usage.ipynb — Demonstrates how to use the software in this repository.
• 📷 phantoms.ipynb — Details for the data used in the paper (Teledyne DALSA Xineos-3131 detectors), available at Zenodo DOI: 10.5281/zenodo.15383254.
• 🫘 2detect.ipynb — Contains the kernel estimation procedure for the 2DeTeCT dataset (Nature Scientific Data DOI: 10.1038/s41597-023-02484-6).

---

📚 Citation

If you use this code or build upon this work, please cite:

@article{Graas_2025,
  doi = {10.1088/1361-6501/addc06},
  url = {https://dx.doi.org/10.1088/1361-6501/addc06},
  year = {2025},
  month = {jun},
  publisher = {IOP Publishing},
  volume = {36},
  number = {6},
  pages = {065415},
  author = {Graas, Adriaan and Lucka, Felix},
  title = {Scintillator decorrelation for self-supervised {X}-ray radiograph denoising},
  journal = {Measurement Science and Technology}
}