find_peaks

Finds local peaks in the distance transform. This is need for watershed segmentations.

import numpy as np
import porespy as ps
import scipy.ndimage as spim
import matplotlib.pyplot as plt
from edt import edt
import inspect
np.random.seed(11)
ps.visualization.set_mpl_style()
inspect.signature(ps.filters.find_peaks)

<Signature (dt, r_max=4, strel=None, sigma=None, divs=1)>

dt

The distance transform must be provided

im = ps.generators.blobs(shape=[100, 100], blobiness=0.5, porosity=0.7)
dt = edt(im)

fig, ax = plt.subplots(1, 2, figsize=[6, 3])

pk = ps.filters.find_peaks(dt=dt)
ax[0].imshow(dt/im)
ax[0].axis(False)

ax[1].imshow(dt/im/~pk)
ax[1].axis(False);

r_max

The radius that should be searched for local maxima.

fig, ax = plt.subplots(1, 2, figsize=[6, 3])

r_max = 3
pk = ps.filters.find_peaks(dt=dt, r_max=r_max)
ax[0].imshow(pk/im)
ax[0].axis(False)
ax[0].set_title(f'r_max = {r_max}')

r_max = 7
pk = ps.filters.find_peaks(dt=dt, r_max=r_max)
ax[1].imshow(pk/im)
ax[1].axis(False)
ax[1].set_title(f'r_max = {r_max}');

strel

The structuring element to use, in the form of a function handle. The value of r_max is passed to this function, so note that the meaning for r differs depending on the strel used. The default is a disk/ball.

im = ps.generators.blobs(shape=[100, 100], blobiness=0.5, porosity=0.7)
dt = edt(im)

from skimage.morphology import disk, square

fig, ax = plt.subplots(1, 2, figsize=[6, 3])

pk = ps.filters.find_peaks(dt=dt, strel=disk(3))
ax[0].imshow(pk/im)
ax[0].axis(False)
ax[0].set_title('disk')

pk = ps.filters.find_peaks(dt=dt, strel=square(5))
ax[1].imshow(pk/im)
ax[1].axis(False)
ax[1].set_title('square');