Creating Multiscale Images

Sometimes it is useful or necessary to study processes across multiple scales. It is possible to generate artificial images with 2 scales of porosity fairly easily.

import porespy as ps
import matplotlib.pyplot as plt

Using blobs of different blobiness

Start by creating one image at a large scale. Let’s use blobs to define the large scale ‘super-structure’:

im1 = ps.generators.blobs([500, 500], blobiness=0.5, porosity=0.6)
fig, ax = plt.subplots(1, 1, figsize=[6, 6])
ax.imshow(im1, interpolation='none')
ax.axis(False);

Now we’ll can create a second image of blobs, but with a different scale. These will represent the ‘sub-structure’:

im2 = ps.generators.blobs([500, 500], blobiness=2.5, porosity=0.5)
fig, ax = plt.subplots(1, 1, figsize=[6, 6])
ax.imshow(im2, interpolation='none')
ax.axis(False);

Finally, we can multiply these two image (i.e. arrays) together, which will have the effect of perforating the foreground phase of im1 with holes defined by the background phase of im2. Note the inverting of the image to ensure the foreground and background are maintained:

im3 = ~(~im1*im2)
fig, ax = plt.subplots(1, 1, figsize=[6, 6])
ax.imshow(im3, interpolation='none')
ax.axis(False);

The pore sizes can be analyzed using local_thickess, to confirm a bi-modal pore size distribution:

lt = ps.filters.local_thickness(im3)
psd = ps.metrics.pore_size_distribution(lt)

ps.visualization.bar(psd);

Using rsa to insert non-overlapping spheres into blobs

Another way is to use rsa to insert spheres into the background of another image. Let’s use blobs to define the superstructure, the perforate it with holes:

im4 = ps.generators.rsa(im1, r=10, clearance=-2)
fig, ax = plt.subplots(1, 1, figsize=[6, 6])
ax.imshow(im4, interpolation='none')
ax.axis(False);

/tmp/ipykernel_8420/1658317612.py:1: DeprecationWarning: Call to deprecated function (or staticmethod) rsa. (This function will be renamed random_spheres in a future version)
  im4 = ps.generators.rsa(im1, r=10, clearance=-2)

Summary

The concept of generate multiple images and different scales then superimposing them is a very easy to make some simple multiscale images for prototyping and testing workflows that require multiscale images.