sknw

Analysis of object skeletons


License
BSD-3-Clause
Install
pip install sknw==0.12

Documentation

Skeleton Network

build net work from nd skeleton image

graph = sknw.build_sknw(ske, multi=False, iso=True, ring=True, full=True)

ske: should be a nd skeleton image

multi: if True,a multigraph is retured, which allows more than one edge between two nodes and self-self edge.

iso: if True, return one-pixel node

ring: if True, return ring without any branch (and insert a self-connected node in the ring)

full: if True, every edge start from the node's centroid. else touch the node block but not from the centroid.

return: is a networkx Graph object

image

graph detail:

graph.nodes[id]['pts'] : Numpy(x, n), coordinates of nodes points

graph.nodes[id]['o']: Numpy(n), centroid of the node

graph.edge(id1, id2)['pts']: Numpy(x, n), sequence of the edge point

graph.edge(id1, id2)['weight']: float, length of this edge

if it's a multigraph, you must add a index after two node id to get the edge, like: graph.edge(id1, id2)[0].

Build Graph:

build Graph by Skeleton, then plot as a vector Graph in matplotlib.

from skimage.morphology import skeletonize
from skimage import data
import sknw

# open and skeletonize
img = data.horse()
ske = skeletonize(~img).astype(np.uint16)

# build graph from skeleton
graph = sknw.build_sknw(ske)

# draw image
plt.imshow(img, cmap='gray')

# draw edges by pts
for (s,e) in graph.edges():
    ps = graph[s][e]['pts']
    plt.plot(ps[:,1], ps[:,0], 'green')
    
# draw node by o
nodes = graph.nodes()
ps = np.array([nodes[i]['o'] for i in nodes])
plt.plot(ps[:,1], ps[:,0], 'r.')

# title and show
plt.title('Build Graph')
plt.show()

Find Path

then you can use networkx do what you want

3D Skeleton

sknw can works on nd image, this is a 3d demo by mayavi

About ImagePy

https://github.com/Image-Py/imagepy

ImagePy is my opensource image processihng framework. It is the ImageJ of Python, you can wrap any numpy based function esaily. And sknw is a sub module of ImagePy. You can use sknw without any code.