-
Notifications
You must be signed in to change notification settings - Fork 0
/
photoMontage3.pyx
171 lines (128 loc) · 5.82 KB
/
photoMontage3.pyx
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
import numpy as np
import maxflow
import sys
import math
cimport cython
@cython.boundscheck(False) # Deactivate bounds checking
@cython.wraparound(False) # Deactivate negative indexing.
cdef inline int color_diff(int[:,:,:] img_a, int[:, :,:] img_b, int x, int y):
cdef int d = 0
cdef int c
for c in range(3):
d += (img_a[y, x, c] - img_b[y, x, c]) ** 2
return math.sqrt(d)
@cython.boundscheck(False) # Deactivate bounds checking
@cython.wraparound(False) # Deactivate negative indexing.
def solve(int [:,:,:,:] photos, mask):
cdef double e = sys.float_info.epsilon
assignment = np.zeros(mask.shape, dtype=np.intc)
assignment[mask != -1] = mask[ mask != -1]
cdef Py_ssize_t rows = mask.shape[0]
cdef Py_ssize_t cols = mask.shape[1]
cdef int [:,:]mask_view = mask
cdef int [:,:]assignment_view
graph = maxflow.Graph[float]()
converged = False
cdef int x,y, u, v, label_u, label_v, var_idx
cdef double var_a, var_b, var_c, var_d, delta, energy
cdef double inf = np.inf
cdef double min_energy = inf
while not converged:
assignment_view = assignment
converged = True
for alpha in range(4):
print("At alpha %d", alpha)
graph.reset()
nodeids = graph.add_grid_nodes(rows * cols)
# sink_edges = np.zeros((rows, cols))
# sink_edges[(mask != -1) & (mask != alpha)] = np.inf
# graph.add_grid_tedges(nodeids, 0, sink_edges.flatten())
var_idx = 0
for y in range(rows):
for x in range(cols):
if mask_view[y, x] != -1 and mask_view[y, x] != alpha:
graph.add_tedge(var_idx, 0, inf)
var_idx += 1
# u = np.arange(1, rows * (cols -1) + 1).reshape((rows, cols - 1))
# v = u - 1
#
# label_u = assignment[:, 1:]
# label_v = assignment[:, :-1]
# Convert loops to vectorised operations
for y in range(rows):
for x in range(1, cols):
u = y * cols + x
v = u - 1
label_u = assignment_view[y,x]
label_v = assignment_view[y, x-1]
if label_u == alpha and label_v == alpha:
continue
var_a = 0.0
var_b = color_diff(photos[alpha], photos[label_v], x,y) + \
color_diff(photos[alpha], photos[label_v], x-1,y)
var_c = color_diff(photos[label_u], photos[alpha], x,y) + \
color_diff(photos[label_u], photos[alpha], x-1,y)
var_d = color_diff(photos[label_u], photos[label_v], x,y) + \
color_diff(photos[label_u], photos[label_v], x-1,y)
if var_a + var_d > var_c + var_b:
delta = var_a + var_d - var_c -var_b
var_a -= delta/3 - e
var_c += delta/3 + e
var_b = var_a + var_d -var_c + e
graph.add_tedge(u, var_d, var_a)
var_b -= var_a
var_c -= var_d
var_b += e
var_c += e
if var_b < 0:
graph.add_tedge(u, 0, var_b)
graph.add_tedge(v, 0, -var_b)
graph.add_edge(u, v, 0.0, var_b + var_c)
elif var_c < 0:
graph.add_tedge(u, 0, -var_c)
graph.add_tedge(v, 0, var_c)
graph.add_edge(u, v, var_b + var_c, 0.0)
else:
graph.add_edge(u, v, var_b, var_c)
for y in range(1, rows):
for x in range(cols):
u = y * cols + x
v = (y-1) * cols + x
label_u = assignment_view[y,x]
label_v = assignment_view[y - 1, x]
if label_u == alpha and label_v == alpha:
continue
var_a = 0.0
var_b = color_diff(photos[alpha], photos[label_v],x,y) + \
color_diff(photos[alpha], photos[label_v], x, y-1)
var_c = color_diff(photos[label_u], photos[alpha], x,y) + \
color_diff(photos[label_u], photos[alpha], x, y-1)
var_d = color_diff(photos[label_u], photos[label_v], x,y) + \
color_diff(photos[label_u], photos[label_v], x, y-1)
if var_a + var_d > var_c + var_b:
delta = var_a + var_d - var_c -var_b
var_a -= delta/3 - e
var_c += delta/3 + e
var_b = var_a + var_d -var_c + e
graph.add_tedge(u, var_d, var_a)
var_b -= var_a
var_c -= var_d
var_b += e
var_c += e
if var_b < 0:
graph.add_tedge(u, 0, var_b)
graph.add_tedge(v, 0, -var_b)
graph.add_edge(u, v, 0.0, var_b + var_c)
elif var_c < 0:
graph.add_tedge(u, 0, -var_c)
graph.add_tedge(v, 0, var_c)
graph.add_edge(u, v, var_b + var_c, 0.0)
else:
graph.add_edge(u, v, var_b, var_c)
energy = graph.maxflow()
if energy < min_energy:
min_energy = energy
converged = False
sgm = graph.get_grid_segments(nodeids)
assignment[~sgm.reshape(assignment.shape)] = alpha
return assignment