凸包

二值图像的凸包是包含输入中所有白色像素的最小凸多边形所包含的像素集合。

该算法的良好概述可以在 Steve Eddin 的博客 上找到。

import matplotlib.pyplot as plt

from skimage.morphology import convex_hull_image
from skimage import data, img_as_float
from skimage.util import invert

# The original image is inverted as the object must be white.
image = invert(data.horse())

chull = convex_hull_image(image)

fig, axes = plt.subplots(1, 2, figsize=(8, 4))
ax = axes.ravel()

ax[0].set_title('Original picture')
ax[0].imshow(image, cmap=plt.cm.gray)
ax[0].set_axis_off()

ax[1].set_title('Transformed picture')
ax[1].imshow(chull, cmap=plt.cm.gray)
ax[1].set_axis_off()

plt.tight_layout()
plt.show()

我们准备第二个图来展示差异。

chull_diff = img_as_float(chull.copy())
chull_diff[image] = 2

fig, ax = plt.subplots()
ax.imshow(chull_diff, cmap=plt.cm.gray)
ax.set_title('Difference')
plt.show()