-
Notifications
You must be signed in to change notification settings - Fork 2.4k
/
024-random_walker_segmentation_scikit-image.py
83 lines (52 loc) · 2.69 KB
/
024-random_walker_segmentation_scikit-image.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
#!/usr/bin/env python
__author__ = "Sreenivas Bhattiprolu"
__license__ = "Feel free to copy, I appreciate if you acknowledge Python for Microscopists"
# https://www.youtube.com/watch?v=6P8YhJa2V6o
"""
Using Random walker to generate lables and then segment and finally cleanup using closing operation.
"""
import matplotlib.pyplot as plt
from skimage import io, img_as_float
import numpy as np
img = img_as_float(io.imread("images/Alloy_noisy.jpg"))
#plt.hist(img.flat, bins=100, range=(0, 1))
# Very noisy image so histogram looks horrible. Let us denoise and see if it helps.
from skimage.restoration import denoise_nl_means, estimate_sigma
sigma_est = np.mean(estimate_sigma(img, multichannel=True))
denoise_img = denoise_nl_means(img, h=1.15 * sigma_est, fast_mode=True,
patch_size=5, patch_distance=3, multichannel=True)
#plt.hist(denoise_img.flat, bins=100, range=(0, 1))
# Much better histogram and now we can see two separate peaks.
#Still close enough so cannot use histogram based segmentation.
#Let us see if we can get any better by some preprocessing.
#Let's try histogram equalization
from skimage import exposure #Contains functions for hist. equalization
#eq_img = exposure.equalize_hist(denoise_img)
eq_img = exposure.equalize_adapthist(denoise_img)
#plt.imshow(eq_img, cmap='gray')
#plt.hist(denoise_img.flat, bins=100, range=(0., 1))
#Not any better. Let us stretch the hoistogram between 0.7 and 0.95
# The range of the binary image spans over (0, 1).
# For markers, let us include all between each peak.
markers = np.zeros(img.shape, dtype=np.uint)
markers[(eq_img < 0.8) & (eq_img > 0.7)] = 1
markers[(eq_img > 0.85) & (eq_img < 0.99)] = 2
from skimage.segmentation import random_walker
# Run random walker algorithm
# https://scikit-image.org/docs/dev/api/skimage.segmentation.html#skimage.segmentation.random_walker
labels = random_walker(eq_img, markers, beta=10, mode='bf')
plt.imsave("images/markers.jpg", markers)
segm1 = (labels == 1)
segm2 = (labels == 2)
all_segments = np.zeros((eq_img.shape[0], eq_img.shape[1], 3)) #nothing but denoise img size but blank
all_segments[segm1] = (1,0,0)
all_segments[segm2] = (0,1,0)
#plt.imshow(all_segments)
from scipy import ndimage as nd
segm1_closed = nd.binary_closing(segm1, np.ones((3,3)))
segm2_closed = nd.binary_closing(segm2, np.ones((3,3)))
all_segments_cleaned = np.zeros((eq_img.shape[0], eq_img.shape[1], 3))
all_segments_cleaned[segm1_closed] = (1,0,0)
all_segments_cleaned[segm2_closed] = (0,1,0)
plt.imshow(all_segments_cleaned)
plt.imsave("images/random_walker.jpg", all_segments_cleaned)