-
Notifications
You must be signed in to change notification settings - Fork 1
/
BAMSFE_BAMS.py
147 lines (133 loc) · 6.5 KB
/
BAMSFE_BAMS.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
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
# -*- coding: UTF-8 -*-
import numpy as np
from sklearn.decomposition import PCA
import matlab.engine
import gc
import matplotlib.pyplot as plt
eng=matlab.engine.start_matlab()
matlab_add_path="addpath(genpath(cd(cd(cd(cd)))))"
eng.eval(matlab_add_path)
def performPCA(img, n_components, eigValuesFlag=False):
if img.ndim == 3:
X = np.reshape(img, (img.shape[0] * img.shape[1], img.shape[-1]))
else:
X = np.reshape(img, (-1, 1))
X_centered = X - np.mean(X, axis=0)
pca = PCA(n_components=n_components)
pca.fit(X_centered)
# Extract eigenvalues
eigenvalues = pca.explained_variance_
X_transformed = pca.transform(X_centered)
img_reconstructed = np.squeeze(np.reshape(X_transformed, (img.shape[0], img.shape[1], -1)))
X = None
X_centered = None
pca = None
X_transformed = None
gc.collect()
if eigValuesFlag:
return np.squeeze(img_reconstructed), eigenvalues[:img_reconstructed.shape[-1]]
else:
return np.squeeze(img_reconstructed)
# @profile()
def BAMSFE(img, supersize_step=5, pca_per_v=0.99, threshold_a=0.01,lambda_v=0.5, sigma=5,pca_flag=True, supersizeReturn=False, args=None, segLabel=False):
if pca_flag:
img_PCA = performPCA(img, n_components=pca_per_v)
else:
img_PCA = img
superpixels_value_dstack = None
if len(img_PCA.shape) == 3:
# print('PCA保留波段数目：{}'.format(img_PCA.shape[-1]))
superpixelsImgValue_list = []
superpixelsLabel_list=[]
supersize_list_all = []
flag = 0
supersize_i = 0
supersize = 0
EV_list = []
while flag < 2:
supersize = supersize + supersize_step
supernum = int(img_PCA.shape[0] * img_PCA.shape[1] / supersize / supersize)
superpixelsLabel = eng.ERS_BAMS(eng.double(img_PCA), eng.double(supernum), eng.double(lambda_v), eng.double(sigma))
superpixelsImgValue = eng.getsuperpixelsvalue_matlab_BAMS(eng.double(superpixelsLabel), eng.double(img_PCA))
EV_v = eng.optimization_matlab_BAMS(eng.double(img_PCA), eng.double(superpixelsLabel))
EV_list.append(EV_v)
if supersize_i > 0:
if (EV_list[supersize_i] / np.sum(EV_list) < threshold_a) or (EV_list[supersize_i] / EV_list[supersize_i - 1] > (1 - threshold_a)):
flag = 2
if flag < 2:
supersize_list_all.append(supersize)
superpixelsImgValue_list.append(superpixelsImgValue)
superpixelsLabel_list.append(superpixelsLabel)
del superpixelsImgValue
supersize_i = supersize_i + 1
del superpixelsLabel
del EV_v
gc.collect()
superpixelsImgValue_list_array = np.array(superpixelsImgValue_list)
superpixelsLabel_list_array=np.array(superpixelsLabel_list)
for supersize_ii, supersize_ in enumerate(supersize_list_all):
if supersize_ii == 0:
superpixels_value_dstack = np.squeeze(superpixelsImgValue_list_array[supersize_ii, :, :, :])
superpixels_label_dstack=np.squeeze(superpixelsLabel_list_array[supersize_ii, :,:,:])
else:
superpixels_value_dstack = np.dstack((superpixels_value_dstack, np.squeeze(superpixelsImgValue_list_array[supersize_ii, :, :, :])))
superpixels_label_dstack=np.dstack((superpixels_label_dstack, np.squeeze(superpixelsLabel_list_array[supersize_ii, :,:,:])))
if args:
plt.plot(np.arange(len(EV_list)), EV_list)
plt.savefig(args.figuresavepath+'EV_list'+args.figure_name+'.svg')
plt.close()
del superpixelsImgValue_list
del EV_list
del superpixelsImgValue_list_array
gc.collect()
img_PCA = None
else:
# print('PCA保留波段数目：1')
img_PCA = np.squeeze(img_PCA)
superpixelsImgValue_list = []
superpixelsLabel_list=[]
supersize_list_all = []
flag = 0
supersize_i = 0
supersize = 0
EV_list = []
while flag < 2:
supersize = supersize + supersize_step
supernum = int(img.shape[0] * img.shape[1] / supersize / supersize)
superpixelsLabel = eng.ERS_BAMS(eng.double(img_PCA), eng.double(supernum), eng.double(lambda_v), eng.double(sigma))
superpixelsImgValue = eng.getsuperpixelsvalue_matlab_BAMS_pro(eng.double(superpixelsLabel), eng.double(img_PCA))
superpixelsImgValue_array = np.array(superpixelsImgValue).reshape(img_PCA.shape)
EV_v = eng.optimization_matlab_BAMS(eng.double(img_PCA), eng.double(superpixelsLabel))
EV_list.append(EV_v)
if supersize_i > 0:
if (EV_list[supersize_i] / np.sum(EV_list) < threshold_a) or (EV_list[supersize_i] / EV_list[supersize_i - 1] > (1 - threshold_a)):
flag = 2
if flag < 2:
supersize_list_all.append(supersize)
superpixelsImgValue_list.append(superpixelsImgValue_array)
superpixelsLabel_list.append(superpixelsLabel)
supersize_i = supersize_i + 1
del superpixelsLabel, superpixelsImgValue_array, EV_v
gc.collect()
superpixelsImgValue_list_array = np.array(superpixelsImgValue_list)
superpixelsLabel_list_array=np.array(superpixelsLabel_list)
for supersize_ii, supersize_ in enumerate(supersize_list_all):
if supersize_ii == 0:
superpixels_value_dstack = np.squeeze(superpixelsImgValue_list_array[supersize_ii, :, :])
superpixels_label_dstack=np.squeeze(superpixelsLabel_list_array[supersize_ii, :,:])
else:
superpixels_value_dstack = np.dstack((superpixels_value_dstack, np.squeeze(superpixelsImgValue_list_array[supersize_ii, :, :])))
superpixels_label_dstack=np.dstack((superpixels_label_dstack, np.squeeze(superpixelsLabel_list_array[supersize_ii, :,:])))
gc.collect()
del superpixelsImgValue_list, EV_list, superpixelsImgValue_list_array
gc.collect()
img_PCA = None
eng.clear(nargout=0)
eng.close()
img_PCA = None
if supersizeReturn:
return superpixels_value_dstack, supersize_list_all
elif segLabel:
return superpixels_label_dstack
else:
return superpixels_value_dstack