apply_padded

Applies padding to an image before passing through specified function

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

[01:01:21] ERROR    PARDISO solver not installed, run `pip install pypardiso`. Otherwise,          _workspace.py:56
                    simulations will be slow. Apple M chips not supported.                                         

im

Generate a test image. Can be 2D or 3D image.

im = np.zeros([300, 300])
im = ps.generators.random_spheres(im=im, r=20, phi=0.6) == 0

plt.figure(figsize=[6, 6])
plt.axis(False)
plt.imshow(im);

func

A good use case is skeletonize_3d, which has edge artifacts that can be reduced if the image is pre-padded. It works on 2D or 3D images.

func = skimage.morphology.skeletonize_3d

pad_width

The amount of padding to add to each axis

x1 = ps.filters.apply_padded(im=im, pad_width=1, func=func, pad_val=1)
x2 = ps.filters.apply_padded(im=im, pad_width=100, func=func, pad_val=1)

fig, ax = plt.subplots(1, 2, figsize=[12, 12]);
ax[0].imshow(x1);
ax[0].axis(False)
ax[0].set_title('pad_width = 1')
ax[1].imshow(x2);
ax[1].axis(False)
ax[1].set_title('pad_width = 100');

/home/runner/work/porespy/porespy/src/porespy/filters/_funcs.py:87: FutureWarning: `skeletonize_3d` is deprecated since version 0.23 and will be removed in version 0.25. Use `skimage.morphology.skeletonize` instead.
  temp = func(padded, **kwargs)

It’s also useful for morphological operations such as binary_opening

func = skimage.morphology.binary_opening

x3 = ps.filters.apply_padded(im=im, pad_width=0, func=func, pad_val=1, footprint=ps.tools.ps_disk(15))
x4 = ps.filters.apply_padded(im=im, pad_width=50, func=func, pad_val=1, footprint=ps.tools.ps_disk(15))

fig, ax = plt.subplots(1, 2, figsize=[12, 12]);
ax[0].imshow(x3)
ax[0].axis(False)
ax[0].set_title('With no padding')
ax[1].imshow(x4)
ax[1].axis(False)
ax[1].set_title('With padding');

pad_val

The value to be placed in padded voxels. It is almost always 1 but setting to 0 may be useful for applying a distance transform.

func = edt

x5 = ps.filters.apply_padded(im=im, pad_width=10, func=func, pad_val=1)
x6 = ps.filters.apply_padded(im=im, pad_width=10, func=func, pad_val=0)

fig, ax = plt.subplots(1, 2, figsize=[12, 12]);
ax[0].imshow(x5);
ax[0].axis(False)
ax[0].set_title('pad_val = 1')
ax[1].imshow(x6);
ax[1].axis(False)
ax[1].set_title('pad_val = 0');