-
Notifications
You must be signed in to change notification settings - Fork 2
/
seam.py
151 lines (111 loc) · 5.21 KB
/
seam.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
import numpy as np
class SeamCarve():
__max_energy = 1000000.0
def __init__(self, img):
self.__arr = img.astype(int)
self.__height, self.__width = img.shape[:2]
self.__energy_arr = np.empty((self.__height, self.__width))
self.__compute_energy_arr()
def __is_border(self, i, j):
return (i == 0 or i == self.__height - 1) or (j == 0 or j == self.__width - 1)
def __compute_energy(self, i, j):
if self.__is_border(i, j):
return self.__max_energy
b = abs(self.__arr[i - 1, j, 0] - self.__arr[i + 1, j, 0])
g = abs(self.__arr[i - 1, j, 1] - self.__arr[i + 1, j, 1])
r = abs(self.__arr[i - 1, j, 2] - self.__arr[i + 1, j, 2])
b += abs(self.__arr[i, j - 1, 0] - self.__arr[i, j + 1, 0])
g += abs(self.__arr[i, j - 1, 1] - self.__arr[i, j + 1, 1])
r += abs(self.__arr[i, j - 1, 2] - self.__arr[i, j + 1, 2])
energy = b + g + r
return energy
def __swapaxes(self):
self.__energy_arr = np.swapaxes(self.__energy_arr, 0, 1)
self.__arr = np.swapaxes(self.__arr, 0, 1)
self.__height, self.__width = self.__width, self.__height
def __compute_energy_arr(self):
self.__energy_arr[[0, -1], :] = self.__max_energy
self.__energy_arr[:, [0, -1]] = self.__max_energy
self.__energy_arr[1:-1, 1:-1] = np.add.reduce(
np.abs(self.__arr[:-2, 1:-1] - self.__arr[2:, 1:-1]), -1)
self.__energy_arr[1:-1, 1:-1] += np.add.reduce(
np.abs(self.__arr[1:-1, :-2] - self.__arr[1:-1, 2:]), -1)
def __compute_seam(self, horizontal=False):
if horizontal:
self.__swapaxes()
energy_sum_arr = np.empty_like(self.__energy_arr)
energy_sum_arr[0] = self.__energy_arr[0]
for i in range(1, self.__height):
energy_sum_arr[i, :-1] = np.minimum(
energy_sum_arr[i - 1, :-1], energy_sum_arr[i - 1, 1:])
energy_sum_arr[i, 1:] = np.minimum(
energy_sum_arr[i, :-1], energy_sum_arr[i - 1, 1:])
energy_sum_arr[i] += self.__energy_arr[i]
seam = np.empty(self.__height, dtype=int)
seam[-1] = np.argmin(energy_sum_arr[-1, :])
seam_energy = energy_sum_arr[-1, seam[-1]]
for i in range(self.__height - 2, -1, -1):
l, r = max(0, seam[i + 1] -
1), min(seam[i + 1] + 2, self.__width)
seam[i] = l + np.argmin(energy_sum_arr[i, l: r])
if horizontal:
self.__swapaxes()
return (seam_energy, seam)
def __carve(self, horizontal=False, seam=None, remove=True):
if horizontal:
self.__swapaxes()
if seam is None:
seam = self.__compute_seam()[1]
if remove:
self.__width -= 1
else:
self.__width += 1
new_arr = np.empty((self.__height, self.__width, 3))
new_energy_arr = np.empty((self.__height, self.__width))
mp_deleted_count = 0
for i, j in enumerate(seam):
if remove:
if self.__energy_arr[i, j] < 0:
mp_deleted_count += 1
new_energy_arr[i] = np.delete(
self.__energy_arr[i], j)
new_arr[i] = np.delete(self.__arr[i], j, 0)
else:
new_energy_arr[i] = np.insert(
self.__energy_arr[i], j, 0, 0)
new_pixel = self.__arr[i, j]
if not self.__is_border(i, j):
new_pixel = (
self.__arr[i, j - 1] + self.__arr[i, j + 1]) // 2
new_arr[i] = np.insert(self.__arr[i], j, new_pixel, 0)
self.__arr = new_arr
self.__energy_arr = new_energy_arr
for i, j in enumerate(seam):
for k in range(j - 1, j + 1):
if 0 <= k < self.__width and self.__energy_arr[i, k] >= 0:
self.__energy_arr[i, k] = self.__compute_energy(i, k)
if horizontal:
self.__swapaxes()
return mp_deleted_count
def resize(self, new_height=None, new_width=None):
if new_height is None:
new_height = self.__height
if new_width is None:
new_width = self.__width
while self.__width != new_width:
self.__carve(horizontal=False, remove=self.__width > new_width)
while self.__height != new_height:
self.__carve(horizontal=True, remove=self.__height > new_height)
def remove_mask(self, mask):
mp_count = np.count_nonzero(mask)
self.__energy_arr[mask] *= -(self.__max_energy ** 2)
self.__energy_arr[mask] -= (self.__max_energy ** 2)
while mp_count:
v_seam_energy, v_seam = self.__compute_seam(False)
h_seam_energy, h_seam = self.__compute_seam(True)
horizontal, seam = False, v_seam
if v_seam_energy > h_seam_energy:
horizontal, seam = True, h_seam
mp_count -= self.__carve(horizontal, seam)
def image(self):
return self.__arr.astype(np.uint8)