-
Notifications
You must be signed in to change notification settings - Fork 4
/
analytical_filters.py
208 lines (174 loc) · 8.17 KB
/
analytical_filters.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
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
#!/usr/bin/env python
u"""
analytical_filters.py
by Yara Mohajerani (Last Update 09/2018)
Use Canny edge detector from sckikit-image to detect edges
Update History
09/2018 written
"""
import os
import sys
import numpy as np
from glob import glob
from PIL import Image, ImageFilter
from skimage import feature
import matplotlib.pyplot as plt
import scipy.misc
from skimage.morphology import skeletonize
from skimage.future import graph
from skimage import data, segmentation, color, filters, io
import sklearn.neighbors
from skimage.filters import sobel
from scipy import ndimage
PLOT = False
#-- read in images
def load_data(suffix,ddir):
#-- initialize dicttionaries
images = {}
files = {}
for d in ['test']:#,'train']:
subdir = os.path.join(ddir[d],'images%s'%(suffix))
#-- get a list of the input files
file_list = glob(os.path.join(subdir,'*.png'))
#-- get just the file names
files[d] = [os.path.basename(i) for i in file_list]
#-- read data
images[d] = {}
#-- get indices of boundaery for each image and select from them randomly
for i,f in enumerate(files[d]):
#-- same file name but different directories for images and labels
images[d][i] = np.array(Image.open(os.path.join(subdir,f)).convert('L'))/255.
return [images,files]
#-- train model and make predictions
def run_filter(parameters):
glacier = parameters['GLACIER_NAME']
suffix = parameters['SUFFIX']
filter = parameters['FILTER']
threshold = np.float(parameters['THRESHOLD'])
#-- directory setup
#- current directory
current_dir = os.path.dirname(os.path.realpath(__file__))
main_dir = os.path.join(current_dir,'..','FrontLearning_data')
glacier_ddir = os.path.join(main_dir,'%s.dir'%glacier)
data_dir = os.path.join(glacier_ddir, 'data')
ddir = {}
ddir['train'] = os.path.join(data_dir,'train')
ddir['test'] = os.path.join(data_dir,'test')
#-- load images
[images,files] = load_data(suffix,ddir)
sigma = 3
#-- go through each image and adjust the sigma until a contiuous front is obtained
for d in ['test']:#,'train']:
if filter =='canny':
#-- make output directory
out_subdir = 'output_canny%s'%suffix
if (not os.path.isdir(os.path.join(ddir[d],out_subdir))):
os.mkdir(os.path.join(ddir[d],out_subdir))
#-- make fronts and save to file
for i in range(len(images[d])):
front = feature.canny(images[d][i], sigma=sigma)
scipy.misc.imsave(os.path.join(ddir[d],out_subdir,'%s'%files[d][i].replace('_Subset',''))\
, front)
if PLOT:
#-- break down image into individual lines by getting indices that are no more than 1 pixel away
#-- from each other. Note 'front' is a boolean array (True / 1 for boundary)
#-- note that the indices are not in order of radial distance so first we need to sort the points one
#-- by one based on distance
indices = np.squeeze(np.nonzero(front)).transpose() # dimensions = num_pts x 2
#-- counter for segments
s = 0
#-- start with the 0 point and add it to the ordered list and then delete
ind = 0
#-- dictionary for indices of segments. Add first point to segment 0
seg = {}
seg[s] = [indices[ind,:]]
indices = np.delete(indices,ind,0)
#-- loop through points until there are no more points left
while len(indices) > 0:
#-- use Ball Tree to search for nearest point to the last point of the current segment
tree = sklearn.neighbors.BallTree(indices, metric='euclidean')
dist,ind = tree.query(seg[s][-1].reshape(1,2), k=1)
#-- assign new point and delete it
if np.squeeze(dist) <=2.:
seg[s].append(indices[ind])
else:
s += 1
seg[s] = [indices[ind]]
indices = np.delete(indices,ind,0)
#-- now plot the longest segment
lens = [len(seg[k]) for k in seg.keys()]
#ind_sorted = np.argsort(lens)
#max_ind = ind_sorted[-2]
max_ind = np.argmax(lens)
new_im = np.zeros(front.shape)
for pix_count in range(lens[max_ind]):
new_im[np.squeeze(np.squeeze(seg[max_ind])[pix_count])[0],\
np.squeeze(np.squeeze(seg[max_ind])[pix_count])[1]] = 1.
# display results
fig, (ax1, ax2, ax3) = plt.subplots(nrows=1, ncols=3, figsize=(8, 3),
sharex=True, sharey=True)
ax1.imshow(images[d][i], cmap=plt.cm.gray)
ax1.axis('off')
ax1.set_title('noisy image', fontsize=20)
ax2.imshow(front, cmap=plt.cm.gray)
ax2.axis('off')
ax2.set_title('Canny filter, $\sigma=%i$'%sigma, fontsize=20)
ax3.imshow(new_im, cmap=plt.cm.gray)
ax3.axis('off')
ax3.set_title('longest line', fontsize=20)
fig.tight_layout()
plt.show()
elif filter == 'sobel':
threshold_str = ''
if threshold != 0:
threshold_str = '_%.2fthreshold'%threshold
#-- make output directory
out_subdir = 'output_sobel%s%s'%(threshold_str,suffix)
if (not os.path.isdir(os.path.join(ddir[d],out_subdir))):
os.mkdir(os.path.join(ddir[d],out_subdir))
#-- make fronts and save to file
for i in range(len(images[d])):
#-- using scikit image sobel filter
front = sobel(images[d][i])
#-- invert image colors
front = 1 - front
if threshold != 0:
#-- set threshold
ind = np.where(front >= threshold)
front[ind] = 1.
#-- using scipy sobel filter
#dx= ndimage.sobel(images[d][i], 0) # horizontal derivative
#dy = ndimage.sobel(images[d][i], 1) # vertical derivative
#front = np.hypot(dx, dy)
#-- using FIL edge detector
#im = Image.fromarray(np.uint8(images[d][i]*255.))
#front = im.filter(ImageFilter.FIND_EDGES)
scipy.misc.imsave(os.path.join(ddir[d],out_subdir,'%s'%files[d][i].replace('_Subset',''))\
, front)
#-- main function to get parameters and pass them along to fitting function
def main():
if (len(sys.argv) == 1):
sys.exit('You need to input at least one parameter file to set run configurations.')
else:
#-- Input Parameter Files (sys.argv[0] is the python code)
input_files = sys.argv[1:]
#-- for each input parameter file
for file in input_files:
#-- keep track of progress
print(os.path.basename(file))
#-- variable with parameter definitions
parameters = {}
#-- Opening parameter file and assigning file ID number (fid)
fid = open(file, 'r')
#-- for each line in the file will extract the parameter (name and value)
for fileline in fid:
#-- Splitting the input line between parameter name and value
part = fileline.split()
#-- filling the parameter definition variable
parameters[part[0]] = part[1]
#-- close the parameter file
fid.close()
#-- pass parameters to training function
run_filter(parameters)
if __name__ == '__main__':
main()