Does medial axis skeletonization work for 3D input arrays ?

[nicobousquet]

Description

Hi, I'm just trying to run the medial axis method in order to skeletonize a 3d object (stl file). But I have this error:

Here is the stl file used in the code:
Turtle_Tree.zip

Way to reproduce

import numpy as np
import matplotlib.pyplot as plt
from scipy import ndimage as ndi
from skimage.morphology import skeletonize, medial_axis
import trimesh
from trimesh.voxel import creation
from mpl_toolkits.mplot3d import Axes3D #don't delete this line

mesh = trimesh.load_mesh('objets 3d/Turtle_Tree.stl')
volume = trimesh.voxel.creation.voxelize(mesh=mesh, pitch=mesh.extents.max() / 128)
array = volume.matrix


filled = ndi.binary_fill_holes(array)
skeleton, distance = medial_axis(filled, return_distance=True)

However, skeletonize_3d works for this input 3d array. So does medial axis work for 3d arrays too or it only does for 2d arrays ?

Thanks in advance.

[jni]

As indicated by the error message, medial axis does not work for 3D images currently.

I don't know much about the medial axis algorithm, but based on my reading of the wikipedia page, the medial axis in a 3D object is in general not a skeleton but a surface. So, although we could (and probably should) update medial axis to work for 3D images, it will not return a skeleton.

[grlee77]

I will go ahead close this as not a bug, but feel free to open a new issue as a feature request for 3D support in skimage.morphology.medial_axis

[DWesl]

feel free to open a new issue as a feature request for 3D support in skimage.morphology.medial_axis

Would #4105 be that feature request?

[nicobousquet]

Personally, I worked on 3D arrays so I would say yes but the other algorithm works pretty well too. I just wanted to compare both methods.