Faster, improved version of UMPA now available on GitHub!

Sample-stepping dataset of a flower
Photo and UMPA-processed image signals of a speckle-based dataset of a small flower, measured with the new sample-stepping technique at P05, PETRA III, DESY. The sample is much larger than the detector field of view (shown as the red rectangle).
We have released a new, improved implementation of UMPA on GitHub! Due to its implementation using C++ and Cython, it is about 120 times faster than the previous Python-based implementation, which is especially useful for large tomographic datasets. Furthermore, we have introduced new features, most prominently “sample-stepping”, which allows displacing the sample instead of the diffuser between acquisitions. This enables a faster acquisition of samples much larger than the field of view. The software is available at: and
An accompanying publication is in review. However, you can already read the preprint at: