Wrong object count in identify_secondary_iw.m

[hackermd]

For some images, the object count is wrong after applying the watershed segmentation. The reason is not clear to me yet. It is related to the shape of the primary objects, however, since smoothing the input mask solves the problem. It could be related to the labeling of objects, which is different in Matlab and Python.

[hackermd]

That was due to differences in labelling of images between Matlab and Python modules.
The mahotas function uses a 4-neighbourhood labelling by default, while Matlab's bwlabel uses a 8-neighbourhood. Fixed with commits 0e8d0bd and 86f6e78.