DSE

Python implementation of discrete skeleton evolution, a skeleton pruning algorithm http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.79.8377&rep=rep1&type=pdf

Setup

pip install -e .

Example This algorithm filters out branches by evalutating their reconstruction weights. Original paper measures the weights by calculating the ratio of reconstruction pixel loss to whole reconstruction pixel, here the weights are simply the reconstruction pixel loss.

from dsepruning import skel_pruning_DSE
import numpy as np
from skimage.io import imread
from skimage.morphology import medial_axis, skeletonize
from scipy.ndimage import distance_transform_edt
import matplotlib
import matplotlib.pyplot as plt

matplotlib.rcParams['figure.figsize'] = [10, 10]

mask = imread('img/skel.png')
mask = mask > 0
plt.imshow(mask)
plt.show()

skel = skeletonize(mask)
print("Show skeleton by skeletonize:")
plt.imshow(skel);plt.show()
dist = distance_transform_edt(mask, return_indices=False, return_distances=True)
print("Show distance map:")
plt.imshow(dist);plt.show()
new_skel = skel_pruning_DSE(skel, dist, 100)
print("Show pruned skeleton")
plt.imshow(new_skel);plt.show()

Show skeleton by skeletonize:

Show distance map:

Show pruned skeleton

skel, dist = medial_axis(mask, return_distance=True)
print("Show skeleton by skeletonize:")
plt.imshow(skel);plt.show()
print("Show distance map:")
plt.imshow(dist);plt.show()
new_skel = skel_pruning_DSE(skel, dist, 100)
print("Show pruned skeleton")
plt.imshow(new_skel);plt.show()

Show skeleton by medial_axis:

Show distance map:

Show pruned skeleton