In [ ]:
%matplotlib inline


# Morphological operations¶

Morphology is the study of shapes. In image processing, some simple operations can get you a long way. The first things to learn are erosion and dilation. In erosion, we look at a pixelâ€™s local neighborhood and replace the value of that pixel with the minimum value of that neighborhood. In dilation, we instead choose the maximum.

In [ ]:
import numpy as np
from matplotlib import pyplot as plt, cm
import skdemo

In [ ]:
image = np.array([[0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0],
[0, 0, 1, 1, 1, 0, 0],
[0, 0, 1, 1, 1, 0, 0],
[0, 0, 1, 1, 1, 0, 0],
[0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0]], dtype=np.uint8)
plt.imshow(image);


The documentation for scikit-image's morphology module is here.

Importantly, we must use a structuring element, which defines the local neighborhood of each pixel. To get every neighbor (up, down, left, right, and diagonals), use morphology.square; to avoid diagonals, use morphology.diamond:

In [ ]:
from skimage import morphology
sq = morphology.square(width=3)
print(sq)
print(dia)


The central value of the structuring element represents the pixel being considered, and the surrounding values are the neighbors: a 1 value means that pixel counts as a neighbor, while a 0 value does not. So:

In [ ]:
skdemo.imshow_all(image, morphology.erosion(image, sq), shape=(1, 2))


and

In [ ]:
skdemo.imshow_all(image, morphology.dilation(image, sq))


and

In [ ]:
skdemo.imshow_all(image, morphology.dilation(image, dia))


Erosion and dilation can be combined into two slightly more sophisticated operations, opening and closing. Here's an example:

In [ ]:
image = np.array([[0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 1, 1, 1, 1, 1, 1, 0, 0],
[0, 0, 1, 1, 1, 1, 1, 1, 0, 0],
[0, 0, 1, 1, 1, 1, 1, 1, 0, 0],
[0, 0, 1, 1, 1, 0, 0, 1, 0, 0],
[0, 0, 1, 1, 1, 0, 0, 1, 0, 0],
[0, 0, 1, 1, 1, 0, 0, 1, 0, 0],
[0, 0, 1, 1, 1, 1, 1, 1, 0, 0],
[0, 0, 1, 1, 1, 1, 1, 1, 0, 0],
[0, 0, 1, 1, 1, 1, 1, 1, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0],
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0]], np.uint8)
plt.imshow(image);


What happens when run an erosion followed by a dilation of this image?

Try to imagine the operations in your head before trying them out below.

In [ ]:
skdemo.imshow_all(image, morphology.opening(image, sq)) # erosion -> dilation

In [ ]:
skdemo.imshow_all(image, morphology.closing(image, sq)) # dilation -> erosion


Exercise: use morphological operations to remove noise from a binary image.

In [ ]:
from skimage import data, color
hub = color.rgb2gray(data.hubble_deep_field()[350:450, 90:190])
plt.imshow(hub);


Remove the smaller objects to retrieve the large galaxy.

In [ ]:
%reload_ext load_style