Signal separation

The core idea is to separate Bragg-peaks from amorphous signal. Peaks are always seen as positive outliers on top of background noise or amorphous signal. The calculation of the background has gradually improved over time:

Initially, it was a median filter on the azimuthal direction after a 2D intergation. The obtained background curve was often noisy. This median filter has been replaced by a sigma-clipping to enforce a normal distribution of the background. Finally the sigma-clipping has been implemented using 1D integrators, avoiding pixel splitting and mixing together rings.

The reconstitution of the Wilson plot is also presented in another notebook.

Finally an alternative implementation of the peakfinder8 algorithm from the Cheetah software is presented.

• Signal separation between amorphous and crystalline phases
• Wilson plots generated from sparse datasets
• Implementation of PeakFinder8 on GPU