distance_transform_lin

A variant of the standard distance transform where the distances are computed along a give axis rather than radially.

import numpy as np
import porespy as ps
import scipy.ndimage as spim
import matplotlib.pyplot as plt
ps.visualization.set_mpl_style()

[19:53:23] ERROR    PARDISO solver not installed, run `pip install pypardiso`. Otherwise,          _workspace.py:56
                    simulations will be slow. Apple M chips not supported.                                         

No module named 'pyedt'

The arguments and their defaults are:

import inspect
inspect.signature(ps.filters.distance_transform_lin)

<Signature (im: numpy.ndarray[typing.Any, numpy.dtype[+_ScalarType_co]], axis: int = 0, mode: Literal['forward', 'backward', 'both'] = 'both')>

axis

The axis along which the distances should be computed

fig, ax = plt.subplots(1, 2, figsize=[12, 6])

im = ps.generators.blobs(shape=[500, 500], porosity=0.7)

axis = 0
dt = ps.filters.distance_transform_lin(im, axis=axis)

ax[0].imshow(dt/im)
ax[0].axis(False)
ax[0].set_title(f'axis = {axis}')

axis = 1
dt = ps.filters.distance_transform_lin(im, axis=axis)

ax[1].imshow(dt/im)
ax[1].axis(False)
ax[1].set_title(f'axis = {axis}');

mode

Whether the distances are comptuted from the start to end, end to start, or both.

fig, ax = plt.subplots(1, 3, figsize=[15, 5])

im = ps.generators.blobs(shape=[500, 500], porosity=0.7)

mode = 'forward'
dt = ps.filters.distance_transform_lin(im, mode=mode)

ax[0].imshow(dt/im)
ax[0].axis(False)
ax[0].set_title(f'mode = {mode}')

mode = 'reverse'
dt = ps.filters.distance_transform_lin(im, mode=mode)

ax[1].imshow(dt/im)
ax[1].axis(False)
ax[1].set_title(f'mode = {mode}')

mode = 'both'
dt = ps.filters.distance_transform_lin(im, mode=mode)

ax[2].imshow(dt/im)
ax[2].axis(False)
ax[2].set_title(f'mode = {mode}');