forked from afrozalm/Patch-Based-Texture-Synthesis
-
Notifications
You must be signed in to change notification settings - Fork 0
/
PatchBasedSynthesis.py
263 lines (241 loc) · 12.7 KB
/
PatchBasedSynthesis.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
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
#!/usr/bin/python
import cv2
import sys
import numpy as np
from random import randint
#Image Loading and initializations
InputName = str(sys.argv[1])
img_sample = cv2.imread(InputName)
img_height = 150
img_width = 150
sample_width = img_sample.shape[1]
sample_height = img_sample.shape[0]
img = np.zeros((img_height,img_width,3), np.uint8)
PatchSize = 30
OverlapWidth = 4
#Picking random patch to begin
randomPatchHeight = randint(0, sample_height - PatchSize)
randomPatchWidth = randint(0, sample_width - PatchSize)
for i in range(PatchSize):
for j in range(PatchSize):
img[i, j] = img_sample[randomPatchHeight + i, randomPatchWidth + j]
#initializating next
GrowPatchLocation = (0,PatchSize)
#---------------------------------------------------------------------------------------#
#| Best Fit Patch and related functions |#
#---------------------------------------------------------------------------------------#
def OverlapErrorVertical( imgPx, samplePx ):
iLeft,jLeft = imgPx
iRight,jRight = samplePx
OverlapErr = 0
diff = np.zeros((3))
for i in range( PatchSize ):
for j in range( OverlapWidth ):
diff[0] = int(img[i + iLeft, j+ jLeft][0]) - int(img_sample[i + iRight, j + jRight][0])
diff[1] = int(img[i + iLeft, j+ jLeft][1]) - int(img_sample[i + iRight, j + jRight][1])
diff[2] = int(img[i + iLeft, j+ jLeft][2]) - int(img_sample[i + iRight, j + jRight][2])
OverlapErr += (diff[0]**2 + diff[1]**2 + diff[2]**2)**0.5
return OverlapErr
def OverlapErrorHorizntl( leftPx, rightPx ):
iLeft,jLeft = leftPx
iRight,jRight = rightPx
OverlapErr = 0
diff = np.zeros((3))
for i in range( OverlapWidth ):
for j in range( PatchSize ):
diff[0] = int(img[i + iLeft, j+ jLeft][0]) - int(img_sample[i + iRight, j + jRight][0])
diff[1] = int(img[i + iLeft, j+ jLeft][1]) - int(img_sample[i + iRight, j + jRight][1])
diff[2] = int(img[i + iLeft, j+ jLeft][2]) - int(img_sample[i + iRight, j + jRight][2])
OverlapErr += (diff[0]**2 + diff[1]**2 + diff[2]**2)**0.5
return OverlapErr
def GetBestPatches( px ):#Will get called in GrowImage
PixelList = []
#check for top layer
if px[0] == 0:
for i in range(sample_height - PatchSize):
for j in range(OverlapWidth, sample_width - PatchSize ):
error = OverlapErrorVertical( (px[0], px[1] - OverlapWidth), (i, j - OverlapWidth) )
if error < ThresholdOverlapError:
PixelList.append((i,j))
elif error < ThresholdOverlapError/2:
return [(i,j)]
#check for leftmost layer
elif px[1] == 0:
for i in range(OverlapWidth, sample_height - PatchSize ):
for j in range(sample_width - PatchSize):
error = OverlapErrorHorizntl( (px[0] - OverlapWidth, px[1]), (i - OverlapWidth, j) )
if error < ThresholdOverlapError:
PixelList.append((i,j))
elif error < ThresholdOverlapError/2:
return [(i,j)]
#for pixel placed inside
else:
for i in range(OverlapWidth, sample_height - PatchSize):
for j in range(OverlapWidth, sample_width - PatchSize):
error_Vertical = OverlapErrorVertical( (px[0], px[1] - OverlapWidth), (i,j - OverlapWidth) )
error_Horizntl = OverlapErrorHorizntl( (px[0] - OverlapWidth, px[1]), (i - OverlapWidth,j) )
if error_Vertical < ThresholdOverlapError and error_Horizntl < ThresholdOverlapError:
PixelList.append((i,j))
elif error_Vertical < ThresholdOverlapError/2 and error_Horizntl < ThresholdOverlapError/2:
return [(i,j)]
return PixelList
#-----------------------------------------------------------------------------------------------#
#| Quilting and related Functions |#
#-----------------------------------------------------------------------------------------------#
def SSD_Error( offset, imgPx, samplePx ):
err_r = int(img[imgPx[0] + offset[0], imgPx[1] + offset[1]][0]) -int(img_sample[samplePx[0] + offset[0], samplePx[1] + offset[1]][0])
err_g = int(img[imgPx[0] + offset[0], imgPx[1] + offset[1]][1]) - int(img_sample[samplePx[0] + offset[0], samplePx[1] + offset[1]][1])
err_b = int(img[imgPx[0] + offset[0], imgPx[1] + offset[1]][2]) - int(img_sample[samplePx[0] + offset[0], samplePx[1] + offset[1]][2])
return (err_r**2 + err_g**2 + err_b**2)/3.0
#---------------------------------------------------------------#
#| Calculating Cost |#
#---------------------------------------------------------------#
def GetCostVertical(imgPx, samplePx):
Cost = np.zeros((PatchSize, OverlapWidth))
for j in range(OverlapWidth):
for i in range(PatchSize):
if i == PatchSize - 1:
Cost[i,j] = SSD_Error((i ,j - OverlapWidth), imgPx, samplePx)
else:
if j == 0 :
Cost[i,j] = SSD_Error((i , j - OverlapWidth), imgPx, samplePx) + min( SSD_Error((i + 1, j - OverlapWidth), imgPx, samplePx),SSD_Error((i + 1,j + 1 - OverlapWidth), imgPx, samplePx) )
elif j == OverlapWidth - 1:
Cost[i,j] = SSD_Error((i, j - OverlapWidth), imgPx, samplePx) + min( SSD_Error((i + 1, j - OverlapWidth), imgPx, samplePx), SSD_Error((i + 1, j - 1 - OverlapWidth), imgPx, samplePx) )
else:
Cost[i,j] = SSD_Error((i, j -OverlapWidth), imgPx, samplePx) + min(SSD_Error((i + 1, j - OverlapWidth), imgPx, samplePx),SSD_Error((i + 1, j + 1 - OverlapWidth), imgPx, samplePx),SSD_Error((i + 1, j - 1 - OverlapWidth), imgPx, samplePx))
return Cost
def GetCostHorizntl(imgPx, samplePx):
Cost = np.zeros((OverlapWidth, PatchSize))
for i in range( OverlapWidth ):
for j in range( PatchSize ):
if j == PatchSize - 1:
Cost[i,j] = SSD_Error((i - OverlapWidth, j), imgPx, samplePx)
elif i == 0:
Cost[i,j] = SSD_Error((i - OverlapWidth, j), imgPx, samplePx) + min(SSD_Error((i - OverlapWidth, j + 1), imgPx, samplePx), SSD_Error((i + 1 - OverlapWidth, j + 1), imgPx, samplePx))
elif i == OverlapWidth - 1:
Cost[i,j] = SSD_Error((i - OverlapWidth, j), imgPx, samplePx) + min(SSD_Error((i - OverlapWidth, j + 1), imgPx, samplePx), SSD_Error((i - 1 - OverlapWidth, j + 1), imgPx, samplePx))
else:
Cost[i,j] = SSD_Error((i - OverlapWidth, j), imgPx, samplePx) + min(SSD_Error((i - OverlapWidth, j + 1), imgPx, samplePx), SSD_Error((i + 1 - OverlapWidth, j + 1), imgPx, samplePx), SSD_Error((i - 1 - OverlapWidth, j + 1), imgPx, samplePx))
return Cost
#---------------------------------------------------------------#
#| Finding Minimum Cost Path |#
#---------------------------------------------------------------#
def FindMinCostPathVertical(Cost):
Boundary = np.zeros((PatchSize),np.int)
ParentMatrix = np.zeros((PatchSize, OverlapWidth))
for i in range(1, PatchSize):
for j in range(OverlapWidth):
if j == 0:
ParentMatrix[i,j] = j if Cost[i-1,j] < Cost[i-1,j+1] else j+1
elif j == OverlapWidth - 1:
ParentMatrix[i,j] = j if Cost[i-1,j] < Cost[i-1,j-1] else j-1
else:
curr_min = j if Cost[i-1,j] < Cost[i-1,j-1] else j-1
ParentMatrix[i,j] = curr_min if Cost[i-1,curr_min] < Cost[i-1,j+1] else j+1
Cost[i,j] += Cost[i-1, ParentMatrix[i,j]]
minIndex = 0
for j in range(1,OverlapWidth):
minIndex = minIndex if Cost[PatchSize - 1, minIndex] < Cost[PatchSize - 1, j] else j
Boundary[PatchSize-1] = minIndex
for i in range(PatchSize - 1,0,-1):
Boundary[i - 1] = ParentMatrix[i,Boundary[i]]
return Boundary
def FindMinCostPathHorizntl(Cost):
Boundary = np.zeros(( PatchSize),np.int)
ParentMatrix = np.zeros((OverlapWidth, PatchSize))
for j in range(1, PatchSize):
for i in range(OverlapWidth):
if i == 0:
ParentMatrix[i,j] = i if Cost[i,j-1] < Cost[i+1,j-1] else i + 1
elif i == OverlapWidth - 1:
ParentMatrix[i,j] = i if Cost[i,j-1] < Cost[i-1,j-1] else i - 1
else:
curr_min = i if Cost[i,j-1] < Cost[i-1,j-1] else i - 1
ParentMatrix[i,j] = curr_min if Cost[curr_min,j-1] < Cost[i-1,j-1] else i + 1
Cost[i,j] += Cost[ParentMatrix[i,j], j-1]
minIndex = 0
for i in range(1,OverlapWidth):
minIndex = minIndex if Cost[minIndex, PatchSize - 1] < Cost[i, PatchSize - 1] else i
Boundary[PatchSize-1] = minIndex
for j in range(PatchSize - 1,0,-1):
Boundary[j - 1] = ParentMatrix[Boundary[j],j]
return Boundary
#---------------------------------------------------------------#
#| Quilting |#
#---------------------------------------------------------------#
def QuiltVertical(Boundary, imgPx, samplePx):
for i in range(PatchSize):
for j in range(Boundary[i], OverlapWidth):
img[imgPx[0] + i, imgPx[1] - j] = img_sample[ samplePx[0] + i, samplePx[1] - j ]
def QuiltHorizntl(Boundary, imgPx, samplePx):
for j in range(PatchSize):
for i in range(Boundary[j], OverlapWidth):
img[imgPx[0] - i, imgPx[0] + j] = img_sample[samplePx[0] - i, samplePx[1] + j]
def QuiltPatches( imgPx, samplePx ):
#check for top layer
if imgPx[0] == 0:
Cost = GetCostVertical(imgPx, samplePx)
# Getting boundary to stitch
Boundary = FindMinCostPathVertical(Cost)
#Quilting Patches
QuiltVertical(Boundary, imgPx, samplePx)
#check for leftmost layer
elif imgPx[1] == 0:
Cost = GetCostHorizntl(imgPx, samplePx)
#Boundary to stitch
Boundary = FindMinCostPathHorizntl(Cost)
#Quilting Patches
QuiltHorizntl(Boundary, imgPx, samplePx)
#for pixel placed inside
else:
CostVertical = GetCostVertical(imgPx, samplePx)
CostHorizntl = GetCostHorizntl(imgPx, samplePx)
BoundaryVertical = FindMinCostPathVertical(CostVertical)
BoundaryHorizntl = FindMinCostPathHorizntl(CostHorizntl)
QuiltVertical(BoundaryVertical, imgPx, samplePx)
QuiltHorizntl(BoundaryHorizntl, imgPx, samplePx)
#--------------------------------------------------------------------------------------------------------#
# Growing Image Patch-by-patch |#
#--------------------------------------------------------------------------------------------------------#
def FillImage( imgPx, samplePx ):
for i in range(PatchSize):
for j in range(PatchSize):
img[ imgPx[0] + i, imgPx[1] + j ] = img_sample[ samplePx[0] + i, samplePx[1] + j ]
pixelsCompleted = 0
TotalPatches = ( (img_height - 1 )/ PatchSize )*((img_width)/ PatchSize) - 1
sys.stdout.write("Progress : [%-20s] %d%% | PixelsCompleted: %d | ThresholdConstant: --.------" % ('='*(pixelsCompleted*20/TotalPatches), (100*pixelsCompleted)/TotalPatches, pixelsCompleted))
sys.stdout.flush()
while GrowPatchLocation[0] + PatchSize < img_height:
pixelsCompleted += 1
ThresholdConstant = 45.0
#set progress to zer0
progress = 0
while progress == 0:
ThresholdOverlapError = ThresholdConstant * PatchSize * OverlapWidth
#Get Best matches for current pixel
List = GetBestPatches(GrowPatchLocation)
if len(List) > 0:
progress = 1
#Make A random selection from best fit pxls
sampleMatch = List[ randint(0, len(List) - 1) ]
FillImage( GrowPatchLocation, sampleMatch )
#Quilt this with in curr location
QuiltPatches( GrowPatchLocation, sampleMatch )
#upadate cur pixel location
GrowPatchLocation = (GrowPatchLocation[0], GrowPatchLocation[1] + PatchSize)
if GrowPatchLocation[1] + PatchSize > img_width:
GrowPatchLocation = (GrowPatchLocation[0] + PatchSize, 0)
#if not progressed, increse threshold
else:
ThresholdConstant *= 1.1
# print pixelsCompleted, ThresholdConstant
sys.stdout.write('\r')
sys.stdout.write("Progress : [%-20s] %d%% | PixelsCompleted: %d | ThresholdConstant: %f" % ('='*(pixelsCompleted*20/TotalPatches), (100*pixelsCompleted)/TotalPatches, pixelsCompleted, ThresholdConstant))
sys.stdout.flush()
# Displaying Images
cv2.imshow('Sample Texture',img_sample)
cv2.waitKey(0)
cv2.destroyAllWindows()
cv2.imshow('Generated Image',img)
cv2.waitKey(0)
cv2.destroyAllWindows()