SNOW partitioning

The filter is used to partition an image into regions using the SNOW algorithm which stands for the subnetwork of the oversegmented watershed. The steps taken are described in detail in the snow_advanced notebook. We provide a filter function that combines all the steps and it is explored here:

import numpy as np
import porespy as ps
import matplotlib.cm as cm
import matplotlib.pyplot as plt
from skimage.morphology import binary_dilation
ps.visualization.set_mpl_style()
np.random.seed(1)

[01:09:23] ERROR    PARDISO solver not installed, run `pip install pypardiso`. Otherwise,          _workspace.py:56
                    simulations will be slow. Apple M chips not supported.                                         

im = ps.generators.overlapping_spheres([500, 500], r=10, porosity=0.5)
fig, ax = plt.subplots()
ax.imshow(im, origin='lower');

snow_out = ps.filters.snow_partitioning(im, r_max=4, sigma=0.4)
print(snow_out)

――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――
Results of snow_partitioning generated at Mon Apr  7 01:09:25 2025
――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――
im                        Array of size (500, 500)
dt                        Array of size (500, 500)
peaks                     Array of size (500, 500)
regions                   Array of size (500, 500)
――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――――

fig, ax = plt.subplots(2, 2, figsize=[8, 8])
ax[0, 0].imshow(snow_out.im, origin='lower')
ax[0, 1].imshow(snow_out.dt, origin='lower')
dt_peak = snow_out.dt.copy()
peaks_dilated = binary_dilation(snow_out.peaks > 0)
dt_peak[peaks_dilated > 0] = np.nan
ax[1, 0].imshow(dt_peak, origin='lower')
ax[1, 1].imshow(ps.tools.randomize_colors(snow_out.regions), origin='lower')
ax[0, 0].set_title("Binary image");
ax[0, 1].set_title("Distance transform");
ax[1, 0].set_title("Distance transform peaks");
ax[1, 1].set_title("Segmentation");