-
Notifications
You must be signed in to change notification settings - Fork 0
/
pairwise_match_lj.py
215 lines (174 loc) · 8.1 KB
/
pairwise_match_lj.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
209
210
211
212
213
214
215
import sys
from collections import defaultdict
import numpy as np
import os
import h5py
import fast64counter
import time
# from util import Util
import matplotlib.pyplot as plt
from tifffile import imsave
Debug = False
# block1_path, block2_path, direction, halo_size, outblock1_path, outblock2_path = sys.argv[1:]
def pair_match(block1,block2, direction, halo_size):
# block1_path = './data/seg1.h5'
# block2_path = './data/seg2_unique.h5'
# outblock1_path = 'out1.h5'
# outblock2_path = 'out2.h5'
# direction = 1
# halo_size = 10
# plt.imshow(block1[4, :, :])
# plt.figure()
# plt.imshow(block2[0, :, :])
direction = int(direction)
halo_size = int(halo_size)
###############################
# Note: direction indicates the relative position of the blocks (1, 2, 3 =>
# adjacent in X, Y, Z). Block1 is always closer to the 0,0,0 corner of the
# volume.
###############################
###############################
# Note that we are still in matlab hdf5 coordinates, so everything is stored ZYX
###############################
###############################
#Change joining thresholds here
###############################
#Join 1 (less joining)
# auto_join_pixels = 20000 # Join anything above this many pixels overlap
# minoverlap_pixels = 2000 # Consider joining all pairs over this many pixels overlap
# minoverlap_dual_ratio = 0.7 # If both overlaps are above this then join
# minoverlap_single_ratio = 0.9# If either overlap is above this then join
# Join 2 (more joining)
# auto_join_pixels = 1000 # Join anything above this many pixels overlap
# minoverlap_pixels = 300 # Consider joining all pairs over this many pixels overlap
# minoverlap_dual_ratio = 0.5 # If both overlaps are above this then join
# minoverlap_single_ratio = 0.8 # If either overlap is above this then join
auto_join_pixels = 300 # Join anything above this many pixels overlap
minoverlap_pixels = 10 # Consider joining all pairs over this many pixels overlap
minoverlap_dual_ratio = 0.1 # If both overlaps are above this then join
minoverlap_single_ratio = 0.1 # If either overlap is above this then join
# print 'Running pairwise matching', " ".join(sys.argv[1:])
# assert block1.size == block2.size
# append the blocks, and pack them so we can use the fast 64-bit counter
# stacked = np.vstack((block1, block2))
# inverse, packed = np.unique(stacked, return_inverse=True)
# packed = packed.reshape(stacked.shape)
# packed_block1 = packed[:block1.shape[0], :, :]
# packed_block2 = packed[block1.shape[0]:, :, :]
# Adjust for Matlab HDF5 storage order
# direction = 3 - direction
direction = direction - 1
stacked = np.concatenate((block1, block2), axis=direction)
inverse, packed = np.unique(stacked, return_inverse=True)
packed = packed.reshape(stacked.shape)
if direction == 0:
packed_block1 = packed[:block1.shape[0], :, :]
packed_block2 = packed[block1.shape[0]:, :, :]
elif direction == 1:
packed_block1 = packed[:, :block1.shape[1], :]
packed_block2 = packed[:, block1.shape[1]:, :]
else:
packed_block1 = packed[:, :, :block1.shape[2]]
packed_block2 = packed[:, :, block1.shape[2]:]
# extract overlap
lo_block1 = [0, 0, 0]
hi_block1 = [None, None, None]
lo_block2 = [0, 0, 0]
hi_block2 = [None, None, None]
# Adjust overlapping region boundaries for direction
lo_block1[direction] = - 1 * halo_size
hi_block2[direction] = 1 * halo_size
block1_slice = tuple(slice(l, h) for l, h in zip(lo_block1, hi_block1))
block2_slice = tuple(slice(l, h) for l, h in zip(lo_block2, hi_block2))
packed_overlap1 = packed_block1[block1_slice]
packed_overlap2 = packed_block2[block2_slice]
print("block1", block1_slice, packed_overlap1.shape)
print("block2", block2_slice, packed_overlap2.shape)
counter = fast64counter.ValueCountInt64()
counter.add_values_pair32(packed_overlap1.astype(np.int32).ravel(), packed_overlap2.astype(np.int32).ravel())
overlap_labels1, overlap_labels2, overlap_areas = counter.get_counts_pair32()
areacounter = fast64counter.ValueCountInt64()
areacounter.add_values(packed_overlap1.ravel())
areacounter.add_values(packed_overlap2.ravel())
areas = dict(zip(*areacounter.get_counts()))
to_merge = []
to_steal = []
merge_dict = {}
for l1, l2, overlap_area in zip(overlap_labels1, overlap_labels2, overlap_areas):
# if inverse[l2] == 8828 and inverse[l1] == 2193:
# bug = 2
if l1 == 0 or l2 == 0:
continue
if ((overlap_area > auto_join_pixels) or
((overlap_area > minoverlap_pixels) and
((overlap_area > minoverlap_single_ratio * areas[l1]) or
(overlap_area > minoverlap_single_ratio * areas[l2]) or
((overlap_area > minoverlap_dual_ratio * areas[l1]) and
(overlap_area > minoverlap_dual_ratio * areas[l2]))))):
if inverse[l2] in merge_dict:
if overlap_area < merge_dict[inverse[l2]][1]:
continue
if inverse[l1] != inverse[l2]:
# print "Merging segments {0} and {1}.".format(inverse[l1], inverse[l2])
to_merge.append((inverse[l1], inverse[l2]))
merge_dict[inverse[l2]]=(inverse[l1],overlap_area)
else:
# print "Stealing segments {0} and {1}.".format(inverse[l1], inverse[l2])
to_steal.append((overlap_area, l1, l2))
# handle merges by rewriting the inverse
merge_map = dict(reversed(sorted(s)) for s in to_merge)
########lj add###############
for idx, val in enumerate(inverse):
if val in merge_map:
while val in merge_map:
val = merge_map[val]
inverse[idx] = val
# Remap and merge
# out1 = h5py.File(outblock1_path + '_partial', 'w')
# out2 = h5py.File(outblock2_path + '_partial', 'w')
# outblock1 = out1.create_dataset('/labels', block1.shape, block1.dtype, chunks=label_chunks, compression='gzip')
# outblock2 = out2.create_dataset('/labels', block2.shape, block2.dtype, chunks=label_chunks, compression='gzip')
# outblock1[...] = inverse[packed_block1]
# outblock2[...] = inverse[packed_block2]
# Util.view(outblock1[40,:,:],large=True)
# Util.view(outblock2[0,:,:],large=True)
# out_one = h5py('out12.h5','w')
# plt.figure()
# plt.imshow(block2[0,:,:]==8836)
outblock1 = inverse[packed_block1]
outblock2 = inverse[packed_block2]
# Util.view(outblock1[2,:,:],large=True,file='temp3.png')
# Util.view(outblock2[0,:,:],large=True,file='temp4.png')
# plt.figure()
# plt.imshow(outblock2[0,:,:])
# out_one = np.vstack((outblock1, outblock2[halo_size:]))
# out_one = np.vstack((outblock1[:outblock1.shape[0]-1], outblock2))
lo_block1 = [0, 0, 0]
hi_block1 = [None, None, None]
lo_block2 = [0, 0, 0]
hi_block2 = [None, None, None]
############ obtain overlap region ############
lo_block1[direction] = - 1 * halo_size
hi_block2[direction] = 1 * halo_size
block1_slice = tuple(slice(l, h) for l, h in zip(lo_block1, hi_block1))
block2_slice = tuple(slice(l, h) for l, h in zip(lo_block2, hi_block2))
block_overlap1 = outblock1[block1_slice]
block_overlap2 = outblock2[block2_slice]
overlap = np.where(block_overlap1>0, block_overlap1, block_overlap2)
###################################
lo_block1 = [0, 0, 0]
hi_block1 = [None, None, None]
lo_block2 = [0, 0, 0]
hi_block2 = [None, None, None]
lo_block2[direction] = 1 * halo_size
block22_slice = tuple(slice(l, h) for l, h in zip(lo_block2, hi_block2))
hi_block1[direction] = 1 * (block1.shape[direction]-halo_size)
block1_slice = tuple(slice(l, h) for l, h in zip(lo_block1, hi_block1))
out_one = np.concatenate((outblock1[block1_slice], overlap, outblock2[block22_slice]), axis=direction)
# return outblock1, outblock2
return out_one
if __name__ == '__main__':
out = np.load('out.npy')
mask = np.load('mask.npy')
out = pair_match(out, mask, direction=1, halo_size=28)
imsave('out.tif', out)