Possible incorrect count normalisation when using MultiGeometry or MultiGeometryFiber to rotate detector

[P-Mousley]

I have been using the MultiGeometry to analyse a set of images collected at various gamma and delta values, and i have found that the normalisation still leaves a faint level of increased intensity for the overlapping sections shared by multiple images.

To check that pyFAI is doing the normalisation correctly i have used the dummy LaB6 dataset included in the documentation and can replicate the same issue - see attached pdf showing contents from the testing notebook. To simulate real-world data i had to add a constant background to the images, as before any pixel other than the powder signal was basically 0 at something like 1e-38.

the important point is demonstrated in the plot below comparing the summed signal, the counts per qmap pixel and the intensity of the integrate2d result

`

pyFAI_stitching_tests.pdf

`

[kif]

Indeed, this looks like a bug ... with the azimuthal version of multigeometry, this should not occur. I will let Edgar deal with the fiber integrator part.