-
Notifications
You must be signed in to change notification settings - Fork 18
/
data_pro_util.py
100 lines (87 loc) · 4.17 KB
/
data_pro_util.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
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
import numpy as np
def norm_img(img, channel_first=True):
if channel_first:
channel, img_height, img_width = img.shape
img = np.reshape(img, (channel, img_height * img_width)) # (channel, height * width)
max_value = np.max(img, axis=1, keepdims=True) # (channel, 1)
min_value = np.min(img, axis=1, keepdims=True) # (channel, 1)
diff_value = max_value - min_value
nm_img = (img - min_value) / diff_value
nm_img = np.reshape(nm_img, (channel, img_height, img_width))
else:
img_height, img_width, channel = img.shape
img = np.reshape(img, (img_height * img_width, channel)) # (channel, height * width)
max_value = np.max(img, axis=0, keepdims=True) # (channel, 1)
min_value = np.min(img, axis=0, keepdims=True) # (channel, 1)
diff_value = max_value - min_value
nm_img = (img - min_value) / diff_value
nm_img = np.reshape(nm_img, (img_height, img_width, channel))
return nm_img
def norm_img_2(img, channel_first=True):
if channel_first:
channel, img_height, img_width = img.shape
img = np.reshape(img, (channel, img_height * img_width)) # (channel, height * width)
max_value = np.max(img, axis=1, keepdims=True) # (channel, 1)
min_value = np.min(img, axis=1, keepdims=True) # (channel, 1)
diff_value = max_value - min_value
nm_img = 2 * ((img - min_value) / diff_value - 0.5)
nm_img = np.reshape(nm_img, (channel, img_height, img_width))
else:
img_height, img_width, channel = img.shape
img = np.reshape(img, (img_height * img_width, channel)) # (channel, height * width)
max_value = np.max(img, axis=0, keepdims=True) # (channel, 1)
min_value = np.min(img, axis=0, keepdims=True) # (channel, 1)
diff_value = max_value - min_value
nm_img = 2 * ((img - min_value) / diff_value - 0.5)
nm_img = np.reshape(nm_img, (img_height, img_width, channel))
return nm_img
def stad_img(img, channel_first=True):
"""
normalization image
:param channel_first:
:param img: (C, H, W)
:return:
norm_img: (C, H, W)
"""
if channel_first:
channel, img_height, img_width = img.shape
img = np.reshape(img, (channel, img_height * img_width)) # (channel, height * width)
mean = np.mean(img, axis=1, keepdims=True) # (channel, 1)
center = img - mean # (channel, height * width)
var = np.sum(np.power(center, 2), axis=1, keepdims=True) / (img_height * img_width) # (channel, 1)
std = np.sqrt(var) # (channel, 1)
nm_img = center / std # (channel, height * width)
nm_img = np.reshape(nm_img, (channel, img_height, img_width))
else:
img_height, img_width, channel = img.shape
img = np.reshape(img, (img_height * img_width, channel)) # (height * width, channel)
mean = np.mean(img, axis=0, keepdims=True) # (1, channel)
center = img - mean # (height * width, channel)
var = np.sum(np.power(center, 2), axis=0, keepdims=True) / (img_height * img_width) # (1, channel)
std = np.sqrt(var) # (channel, 1)
nm_img = center / std # (channel, height * width)
nm_img = np.reshape(nm_img, (img_height, img_width, channel))
return nm_img
def random_select_samples(img_X, img_Y, n_train, patch_sz):
'''
randomly selecting patches as training sample for KPCA convolution training
'''
height, width, channel = img_X.shape
edge = patch_sz // 2
img_X = np.pad(img_X, ((edge, edge), (edge, edge), (0, 0)), 'constant')
img_Y = np.pad(img_Y, ((edge, edge), (edge, edge), (0, 0)), 'constant')
patch_X = []
patch_Y = []
for i in range(0, height):
for j in range(0, width):
patch_X.append(img_X[i:i + patch_sz, j:j + patch_sz])
patch_Y.append(img_Y[i:i + patch_sz, j:j + patch_sz])
patch_X = np.array(patch_X, np.float32)
patch_Y = np.array(patch_Y, np.float32)
train_idx = np.arange(0, height * width)
np.random.seed()
np.random.shuffle(train_idx)
train_idx = train_idx[0:n_train]
train_X = patch_X[train_idx]
train_Y = patch_Y[train_idx]
return train_X, train_Y