Watershed does not segment my array

[Make42]

Description

The watershed function produces an unexpected result: It seems not to be able to segment my array even though I expect it to.

Way to reproduce

from skimage.segmentation import watershed
import numpy as np
myarray = np.array([
       [ 2.50148667,  9.44388663,  3.44389182,  3.02727774,  2.40108087,  1.61053155,  0.82294687],
       [ 3.31356207, 10.55797911,  4.69412493,  4.41801194,  3.80963218,  2.70408164,  1.35735552],
       [ 3.28787642,  4.59215029,  4.97735914, 11.08993154,  4.7925308 ,  9.6116032 ,  1.79749649],
       [ 2.87281939,  4.15976317,  4.70572275, 10.95797537,  4.73970551,  9.56800396,  1.7571604 ],
       [ 1.98065752,  2.97293194,  3.42830952,  3.56329229,  3.27749856,  2.33718172,  1.1267278 ]])
isLocalMaxArray = np.array([
       [False, False, False, False, False, False, False],
       [False,  True, False, False, False, False, False],
       [False, False, False,  True, False,  True, False],
       [False, False, False, False, False, False, False],
       [False, False, False, False, False, False, False]])
watershed(-myarray, isLocalMaxArray, watershed_line=True)

produces

array([
       [1, 1, 1, 1, 1, 1, 1],
       [1, 1, 1, 1, 1, 1, 1],
       [1, 1, 1, 1, 1, 1, 1],
       [1, 1, 1, 1, 1, 1, 1],
       [1, 1, 1, 1, 1, 1, 1]], dtype=int32)

while I expected

array([
       [1, 1, 0, 3, 0, 2, 2],
       [1, 1, 0, 3, 0, 2, 2],
       [1, 1, 0, 3, 0, 2, 2],
       [1, 1, 0, 3, 0, 2, 2],
       [1, 1, 0, 3, 0, 2, 2]], dtype=int32)

Version information

from __future__ import print_function
import sys; print(sys.version)
import platform; print(platform.platform())
import skimage; print("scikit-image version: {}".format(skimage.__version__))
import numpy; print("numpy version: {}".format(numpy.__version__))

3.8.8 (default, Feb 24 2021, 21:46:12)
[GCC 7.3.0]
Linux-5.4.0-67-generic-x86_64-with-glibc2.10
scikit-image version: 0.17.2
numpy version: 1.19.2

[Make42]

After quite a bit of trying out, I found out, that it is not enough to simply provide if there is a peak at a certain position, but a label that indicates which peaks belong together. This can be done with skimage.measure.label

from skimage import measure
watershed(-myarray, measure.label(isLocalMaxArray, background=0), watershed_line=True)