-
Notifications
You must be signed in to change notification settings - Fork 0
/
seamcarving.py
231 lines (168 loc) · 8.01 KB
/
seamcarving.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
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
# CS6475 - Spring 2022
import numpy as np
import scipy as sp
import cv2
import scipy.signal # option for a 2D convolution library
from matplotlib import pyplot as plt # for optional plots
import copy
'''
----------DEBUGGING HELPER FUNCTIONS----------------
'''
def show_float_image(image):
#This method is used for debugging intermediate results
print(image.shape)
norm_uint8_image = 255*((image-image.min())/(image.max()-image.min()))
plt.imshow(norm_uint8_image,cmap = 'gray')
plt.pause(0)
def print_percentage_complete(current, total):
#This method is used for printing percentage completed while debugging
per = int(((current+1)*100)/total)
print("Percentage Completed: "+str(per),end='\r')
'''
----------------------------------------------------
'''
'''
----------ENERGY COST MATRIX FUNCTIONS----------------
'''
def get_backward_energy_cost_matrix(image):
gray_image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY).astype(np.float64)
Ix = cv2.Sobel(src=gray_image, ddepth=cv2.CV_64F, dx=1, dy=0, ksize=3)
Iy = cv2.Sobel(src=gray_image, ddepth=cv2.CV_64F, dx=0, dy=1, ksize=3)
energy_map = abs(Ix) + abs(Iy)
# energy_map = abs(cv2.Sobel(src=gray_image, ddepth=cv2.CV_64F, dx=1, dy=1, ksize=3))
# show_float_image(energy_map)
M = np.copy(energy_map)
height, width = M.shape
for h in range(1, height):
M[h,1:width-1] += np.min(np.array([np.roll(M[h-1],1), M[h-1], np.roll(M[h-1],-1)]), axis = 0)[1:width-1]
M[h,0] += min(M[h-1,0], M[h-1,1])
M[h,width-1] += min(M[h-1,width-2], M[h-1,width-1])
# M[h,0] += min(M[h-1,0], M[h-1,1])
# for w in range(1, width-1):
# M[h,w] += min(M[h-1,w-1],M[h-1,w], M[h-1,w+1])
# M[h,width-1] += min(M[h-1,width-2], M[h-1,width-1])
# show_float_image(M)
return M
def get_forward_energy_cost_matrix(image):
gray_image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY).astype(np.float64)
M = np.zeros(gray_image.shape)
I1 = np.roll(gray_image, 1, axis=0) #I(i-1,j)
I2 = np.roll(gray_image, 1, axis=1) #I(i,j-1)
I3 = np.roll(gray_image, -1, axis=1) #I(i,j+1)
CU = abs(I3-I2) # |I(i,j+1) - I(i,j-1)|
CL = CU + abs(I1-I2) # |I(i,j+1) - I(i,j-1)| + |I(i-1,j) - I(i,j-1)|
CR = CU + abs(I1-I3) # |I(i,j+1) - I(i,j-1)| + |I(i-1,j) - I(i,j+1)|
for i in range(1, gray_image.shape[0]):
# M[i] = min(M[i-1,j-1]+CL[i,j], M[i-1,j]+CU[i,j], M[i-1,j+1]+CR[i,j] )
M[i] = np.min(np.array([np.roll(M[i-1],1)+CL[i], M[i-1]+CU[i], np.roll(M[i-1],-1)+CR[i]]), axis=0)
# show_float_image(M)
return M
'''
----------------------------------------------------
'''
'''
---------------SEAM FUNCTIONS-----------------------
'''
def get_vertical_seam(cost_matrix):
height,width = cost_matrix.shape
seam_indices = [np.argmin(cost_matrix[height-1])]
index_count = 0
for h in range(height-2,-1,-1):
w = seam_indices[index_count]
if w == 0:
seam_indices.append(np.argmin(cost_matrix[h,0:2]))
elif w == width-1:
seam_indices.append(width - 2 + np.argmin(cost_matrix[h,width-2:width]))
else:
seam_indices.append(w - 1 + np.argmin(cost_matrix[h,w-1:w+2]))
index_count += 1
seam_indices.reverse()
return seam_indices
def paint_vertical_removal_seam(image, seam_indices, index_tracker):
for h in range(image.shape[0]):
w = index_tracker[h,seam_indices[h]]
image[h,w] = [0,0,255]
def remove_vertical_seam(image, seam_indices):
if len(image.shape) == 3:
output = np.zeros([image.shape[0],image.shape[1]-1,image.shape[2]],dtype=image.dtype)
else:
output = np.zeros([image.shape[0],image.shape[1]-1],dtype=image.dtype)
for h in range(image.shape[0]):
w = seam_indices[h]
output[h,:w] = image[h,:w]
output[h,w:] = image[h,w+1:]
return output
def insert_vertical_seams(image, seams_list, insertion_count, redSeams = False):
seams = len(seams_list[0])
height, width, channels = image.shape
seam_insertion_image = np.zeros([height, width+seams*insertion_count, channels])
for h in range(height):
seams_list[h].sort()
seam_insertion_image[h, :seams_list[h][0] + 1] = image[h, :seams_list[h][0] + 1]
for w in range(seams):
index = seams_list[h][w]
insertion_image_index = index + w*insertion_count + 1
if w == seams-1:
pixel = image[h,index]
if redSeams:
pixel = [0,0,255]
seam_insertion_image[h,insertion_image_index:insertion_image_index+insertion_count] = pixel
seam_insertion_image[h,insertion_image_index+insertion_count:] = image[h, index+1:]
else:
pixel = image[h,index]/2.0 + image[h,index+1]/2.0
if redSeams:
pixel = [0,0,255]
seam_insertion_image[h,insertion_image_index:insertion_image_index+insertion_count] = pixel
seam_insertion_image[h,insertion_image_index+insertion_count:seams_list[h][w+1]+(w+1)*insertion_count + 1] = image[h, index+1:seams_list[h][w+1]+1]
# print_percentage_complete(h,height)
return seam_insertion_image.astype(image.dtype)
def image_vertical_seam_removal(image, seams, cost_function, redSeams=False):
seam_removal_image = np.copy(image)
red_seam_image = np.copy(image)
index_tracker = np.array([np.arange(image.shape[1])]*image.shape[0])
for seam in range(seams):
cost_matrix = cost_function(seam_removal_image)
seam_indices = get_vertical_seam(cost_matrix)
seam_removal_image = remove_vertical_seam(seam_removal_image, seam_indices)
if redSeams:
paint_vertical_removal_seam(red_seam_image, seam_indices, index_tracker)
index_tracker = remove_vertical_seam(index_tracker, seam_indices)
# print_percentage_complete(seam,seams)
return seam_removal_image, red_seam_image
def image_vertical_seam_insertion(image, seams, cost_function, insertion_count=1, redSeams=False):
seam_removal_image = np.copy(image)
index_tracker = np.array([np.arange(image.shape[1])]*image.shape[0])
height, width, channels = image.shape
seams_list = [[] for _ in range(height)]
for seam in range(seams):
cost_matrix = cost_function(seam_removal_image)
seam_indices = get_vertical_seam(cost_matrix)
seam_removal_image = remove_vertical_seam(seam_removal_image, seam_indices)
for h in range(height):
x = index_tracker[h,seam_indices[h]]
seams_list[h].append(x)
index_tracker = remove_vertical_seam(index_tracker, seam_indices)
# print_percentage_complete(seam,seams)
seam_insertion_image = insert_vertical_seams(image, seams_list, insertion_count)
if redSeams:
red_seam_image = insert_vertical_seams(image, seams_list, insertion_count, True)
else:
red_seam_image = None
return seam_insertion_image, red_seam_image
'''
----------------------------------------------------
'''
if __name__ == "__main__":
image = cv2.imread("images/test1.png")
result, red_result = image_vertical_seam_removal(image, image.shape[1]//2, get_backward_energy_cost_matrix, True)
cv2.imwrite("images/result1.png", result)
cv2.imwrite("images/red_result1.png", red_result)
result, red_result = image_vertical_seam_removal(image, image.shape[1]//2, get_forward_energy_cost_matrix, True)
cv2.imwrite("images/result2.png", result)
cv2.imwrite("images/red_result2.png", red_result)
result, red_result = image_vertical_seam_insertion(image, image.shape[1]//2, get_backward_energy_cost_matrix, 1, True)
cv2.imwrite("images/insertion-result1.png", result)
cv2.imwrite("images/insertion-red_result1.png", red_result)
result, red_result = image_vertical_seam_insertion(image, image.shape[1]//2, get_forward_energy_cost_matrix, 1, True)
cv2.imwrite("images/insertion-result2.png", result)
cv2.imwrite("images/insertion-red_result2.png", red_result)