porespy.beta.pts_to_voronoi#

porespy.beta.pts_to_voronoi(im, r=0, centroids=True, borders=True)#

Compute a Voronoi tessellation from seed points via watershed.

Each True pixel in im is treated as a seed. An EDT-based watershed grows regions from those seeds to fill the domain. Optionally the seeds are iteratively relaxed to region centroids (Lloyd’s algorithm) before the final tessellation is returned.

Parameters:

  • im (ndarray of bool) – A 2-D boolean image where True pixels mark the Voronoi seed points.

  • r (int, optional) – Number of Lloyd relaxation iterations. At each step the seeds are replaced by the centroids of the current Voronoi regions and the tessellation is recomputed. r=0 (default) uses the original seeds without relaxation.

  • centroids (bool, optional) – If True (default), the centroid of each final Voronoi region is marked with the value -1 in the returned label image.

  • borders (bool, optional) – If True (default), watershed lines (the Voronoi boundaries) are included in the output as pixels with value 0. If False, every pixel is assigned to its nearest region and no boundary pixels exist.

Returns:

ws – A 2-D integer label array the same shape as im. Each Voronoi region carries a unique positive integer label. Boundary pixels are 0 (when borders=True), and centroid pixels are -1 (when centroids=True).

Return type:

ndarray of int

Notes

The watershed is seeded from the labelled connected components of im and grows into the distance-transform of ~im. Each relaxation iteration replaces the current seeds with the centre-of-mass of every labelled region, then calls pts_to_voronoi recursively with r=0.

Examples

>>> import numpy as np
>>> im = np.zeros([200, 200], dtype=bool)
>>> pts = (np.random.rand(50, 2) * 200).astype(int)
>>> im[pts[:, 0], pts[:, 1]] = True
>>> ws = pts_to_voronoi(im, r=2)
On this page