Iso-distance profiler

[anderskaestner]

Have you considered implementing an iso-distance profiler? It is more general purpose

Something like:

import numpy as np
from scipy.ndimage import distance_transform_edt as edt

dist     = edt(bi_level_img) # round off to the needed number of digits
dlist    = np.unique(dist.ravel())

profile = np.zeros_like(dlist)
for idx,d in enumerate(dlist) :
    profile[idx] = dist[dist==d].mean()

This is not the most efficient implementation (you have to scan the image as many times as you have distances), but it is a baseline for the principle.