Image analysis

The aim of this section is the extraction of information from images. We are going to focus on two cases:

• Image structure
• Object recognition

Image structure

The goal is the representation of the contents of an image using simple structures. We focus on one case alone: image segmentation. We encourage the reader to explore other settings, such as quadtree decompositions.

Segmentation is a method to represent an image by partition into multiple objects (segments); each of them sharing some common property.

In the case of binary images, we can accomplish this by a process of labeling, as we have shown in a previous section. Let's revisit that technique with an artificial image composed by 30 random disks placed on a 64 x 64 canvas:

In [1]: import numpy as np, matplotlib.pyplot as plt
In [2]: from skimage.draw import circle
In [3]: image = np.zeros((64, 64)).astype('bool')
In [4]: for k in range(30):
   ...:     x0, y0 = np.random.randint(64, size=(2))
   ...:     image...