Documentation needed for the newer peaksearching / segmentation

[jonwright]

We are missing some documentation since the peaksearching/segmentation was overhauled in recent years.

The traditional code is described here:. There is an incantation to run the traditional peaksearch code described in com_guide with the new hdf5 data formats, as an alternative to the hdf5scan.py shim that is in sandbox:

peaksearch.py -n SiliconWaferXrotyCalib_att10.h5::1.1/measurement/eiger -f 0 -l 3600 -t 10 -p Y -F "" --ndigits=0 -o /tmp/test.spt

This is not obvious and still misses a mask option. For ID03 and finterlaced I guess you need --interlaced and --iflip too.

The newer code in ImageD11/sinograms/lima_segmenter is taking a 2D frame and segmenting to create sparse arrays (e.g. saving pixels in peaks). This is similar to pyFAI's sparsify bragg method.

There are currently two pixel to peak labelling methods in C code, for both sparse and dense images. A simple connected components and and also an uphill assignment (localmaxlabel). These are not exposed in most of the user facing code. There are a few different places where there is code for collecting peak moments from labeled images. There is also code to compare in scipy.ndimage or skimage, etc.

• we miss any function to "fit a peak" into a 2D (or 3D stack). This is mainly interesting for overlapping peaks.

For merging peaks across frames we have:

• 3D: peaksearch.py has peaks merged if there are common pixels overlapping
• 3D: very old code in ImageD11.peakmerger and newer code in frelon_peaksearch that merges based on centroids
• 3D + 4D sinograms: recent code based on overlapping pixels

There is an open issues for the 4D merging code

• SparseFrame vs SparseScan and lmlabel #159 SparseFrame vs SparseScan and lmlabel
• ToDo's for the for sparse pixels codes #213 ToDo's for the sparse pixels codes
• Tag peaks that are touching the detector mask #241 Tag peaks that are touching the detector mask
• Allow no mask to be used with lima_segmenter #291 Allow no mask to be used with lima_segmenter
• Get rid of the shared memory arrays : add a colfile mmap option #265 Get rid of the shared memory arrays : add a colfile mmap option

There are GUI things in the old tkinter code, in ImageD11/silxGuis, and in the existing Orange code in Julia's project for ID06:

https://gitlab.esrf.fr/julia.garriga/3DXRD/-/blob/master/imaged11_julia/peakSearchWidget.py?ref_type=heads