Python/Lab 7 at main · AmFruitY/Python · GitHub

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
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
# -*- coding: utf-8 -*-
"""
Created on Fri Apr 21 11:53:02 2023

@author: fis.aules
"""

# Lab 7: K-means clustering
# Joshua Santuyo Lorenzana

#%% Libraries

import numpy as np

import matplotlib.pyplot as plt
import matplotlib.image as mpimg
from skimage import color
from skimage import metrics
from skimage import data

from sklearn import cluster
from sklearn import utils

#%% Images

irc = mpimg.imread('aeroport_irc.jpg')
rgb = mpimg.imread('aeroport_rgb.jpg')

im = np.stack([rgb[:,:,0], rgb[:,:,1], rgb[:,:,2], irc[:,:,0]], axis =-1)

#%% 7.1 Clustering of aerial images

# reshaping the image

def flatten(img):

    imgR = img[:,:,0]
    imgG = img[:,:,1]
    imgB = img[:,:,2]
    imgI = img[:,:,3]

    # flattening and putting into rows

    img_K = np.stack([imgR.flatten(), imgG.flatten(), imgB.flatten(), imgI.flatten()], axis =-1)

    return img_K

im_K = flatten(im)
# system training

im_KMN = cluster.KMeans(n_clusters = 5, init = 'k-means++', random_state = 0).fit(im_K)

#%% Let's create a new cell just to not use fit again, as it is time-consuming

# # labels

# labels_im = im_KMN.predict(im_K)

# # centroids

# centroids_im = im_KMN.cluster_centers_

# # from 1d-array to 2d-array
# imRes =np.reshape(labels_im, [250, 500])

# # Representing

# plt.figure(1)
# plt.imshow(imRes)

# # Producing the scatterplot for the coordinates

# plt.figure(2)

# plt.subplot(2, 3, 1)
# plt.title('R vs G')
# plt.scatter(im_K[:,0], im_K[:,1], c = labels_im , cmap='jet')

# plt.subplot(2, 3, 2)
# plt.title('R vs B')
# plt.scatter(im_K[:,0], im_K[:,2], c = labels_im , cmap='jet')

# plt.subplot(2, 3, 3)
# plt.title('R vs IR')
# plt.scatter(im_K[:,0], im_K[:,3], c = labels_im , cmap='jet')

# plt.subplot(2, 3, 4)
# plt.title('G vs B')
# plt.scatter(im_K[:,1], im_K[:,2], c = labels_im , cmap='jet')

# plt.subplot(2, 3, 5)
# plt.title('G vs I')
# plt.scatter(im_K[:,1], im_K[:,3], c = labels_im , cmap='jet')

# plt.subplot(2, 3, 6)
# plt.title('B vs I')
# plt.scatter(im_K[:,2], im_K[:,3], c = labels_im , cmap='jet')

# plt.tight_layout()

#%% 7.2 Revisitng indexed color

# for this exercise, we'll be using the astronaut image

im2 = data.astronaut()

imgR = im2[:,:,0]
imgG = im2[:,:,1]
imgB = im2[:,:,2]

# flattening and putting into rows

im2_K = np.stack([imgR.flatten(), imgG.flatten(), imgB.flatten()], axis =-1)

# reducing the number of levels to 215

im2_rdc = 5 * np.uint8(im2_K/6)

# plt.figure(3)
# plt.subplot(1,2,1)
# plt.title('Reduced to 216-level')
# plt.imshow(im_rdc)
# plt.subplot(1,2,2)
# plt.title('OG image')
# plt.imshow(im)

# plt.tight_layout()

# # Image Comparison

# ssim = metrics.structural_similarity(im_rdc, im2, multichannel='True')

# # ??? win_size exceeds image extent.  If the input is a multichannel (color) image, set multichannel=True
# print (' The reduced 216-level image and the original image has a similarity of about ' + str(ssim) + '.')

# K-means the reduced image
# system training

# efficiency

target = utils.shuffle(im2_rdc, random_state = 0)
im2_rdc_KMN = cluster.KMeans(n_clusters = 216, init = 'k-means++', random_state = 0).fit(im2_K)

#%%
labels_im2_rdc = im2_rdc_KMN.predict(im2_K)
centroids_im2_rdc = im2_rdc_KMN.cluster_centers_
im2_rdc_res = np.reshape(labels_im2_rdc, [512,512])

plt.figure(4)
plt.subplot(1,2,1)
plt.title('K-means processed image')
plt.imshow(im2_rdc_res)

# new index
index = np.linspace(0, 215, 216, dtype=np.uint8)
lut =np.uint8(centroids_im2_rdc[index])

plt.subplot(1,2,2)
plt.title('Color indexed image using the centroids of the clusters')
plt.imshow(lut[im2_rdc_res])