Based on the discussion here: https://groups.google.com/forum/#!topic/pythonvision/AVrnueiKKYI
It seems like this label function should be faster than the scipy version. However, I find it to be much slower:
from skimage import morphology from scipy import ndarray as nd
k = np.zeros((4000,4000),dtype=int) k[100,100] = 1 strel = np.ones((3,3))
1 loops, best of 3: 1.46 s per loop
1 loops, best of 3: 355 ms per loop
If you have any insight, I would be very grateful. I am using version 0.8.2 on the versions of python/numpy/cython, etc. included in the most recent edition of Enthought/Canopy.