/
photomosaic.py
221 lines (199 loc) · 7.21 KB
/
photomosaic.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
import json
import numpy as np
import cv2
import math
import sys
import copy
INDEX_PATH = "./index/"
def getAverageColor(image, index, bins):
(h,w,_) = image.shape
histogram = cv2.calcHist([image],[index], None, [bins],[0,bins])
x = 0
for i in range(0,len(histogram)):
x += (int(histogram[i])*i)
return x / (w*h)
def extractFeature(image):
entry = {}
entry["b"] = getAverageColor(image, 0, 256)
entry["g"] = getAverageColor(image, 1, 256)
entry["r"] = getAverageColor(image, 2, 256)
image = cv2.cvtColor(image, cv2.COLOR_BGR2HSV)
entry["h"] = getAverageColor(image, 0, 180)
entry["s"] = getAverageColor(image, 1, 256)
entry["v"] = getAverageColor(image, 2, 256)
return entry
def readIndex():
json_data = open(INDEX_PATH + "histogram.index").read()
return json.loads(json_data)
#changes the input image to a size that is a multiple of the tilesize.
def preparInputImage(path, tileSize):
i = cv2.imread(path)
(h, w, _) = i.shape
i = cv2.resize(i, (w / tileSize * tileSize, h / tileSize * tileSize))
return i
def preparePatch(path, tileSize):
image = cv2.imread(INDEX_PATH + path)
image = cv2.resize(image, (tileSize, tileSize))
return image
def calcDistance(fts1, fts2, vectors):
distance = 0
for vec in vectors:
distance += math.pow(fts1[vec] - fts2[vec], 2)
return math.sqrt(distance)
#copmutes the mean squared error (mes)
def avepix(cc,ff):
g4 = 0
b4 = 0
r4 = 0
(h, w, _) = cc.shape
pix = h * w
for i in range(h):
for j in range(w):
r3= int(ff[i][j][0])-int(cc[i][j][0])
g3 = int(ff[i][j][1])-int(cc[i][j][1])
b3=int(cc[i][j][2])-int(ff[i][j][2])
g4 += g3*g3
b4 += b3*b3
r4 += r3*r3
mesg = g4/pix
mesr = r4/pix
mesb = b4/pix
mes = mesb+mesg+mesr
return mes
cv2.get
def getIndexImage(fts, index, vectors):
minDistance = sys.maxint
imagefile = ""
for item in index:
distance = calcDistance(fts, item, vectors)
if distance < minDistance:
minDistance = distance
imagefile = item["file"]
return imagefile
def processLine(w,h, index, inputImage, tileSize, channels):
for i in range(0, h / tileSize):
for j in range(0, w / tileSize):
roi = inputImage[i * tileSize:(i + 1) * tileSize, j * tileSize:(j + 1) * tileSize]
fts = extractFeature(roi)
patch = preparePatch(getIndexImage(fts, index, channels), tileSize)
inputImage[i * tileSize:(i + 1) * tileSize, j * tileSize:(j + 1) * tileSize] = patch
#cv2.imshow("Progress", inputImage)
#cv2.waitKey(1)
return inputImage
def mosaic(tileSize):
inputImagePath = str('pictures/bigpicture/bigimages/starwars101.jpg')
channels = list(str('rgb'))
index = readIndex()
inputImage = preparInputImage(inputImagePath, tileSize)
(h, w, _) = inputImage.shape
inputImage = cv2.resize(inputImage, (w / tileSize * tileSize, h / tileSize * tileSize))
print inputImage.shape
cc = copy.deepcopy(inputImage)
ff = processLine(w,h, index, inputImage, tileSize, channels)
#cv2.imshow("Progress", ff)
#cv2.waitKey(1000)
ff = blend2(ff,tileSize)
#cv2.imshow("Progress", ff)
#cv2.waitKey(1000)
mes = avepix(cc,ff)
print "Finished processing of image"
return mes, inputImage
#simple alpha blending. http://stackoverflow.com/questions/29106702/blend-overlapping-images-in-python
def blend2(image, tilesize):
for i in range(0,len(image)-tileSize,tileSize):
image1 = image[i+tileSize-2:i+tileSize,:]
image2 = image[i+tileSize:i+tileSize+2,:]
alpha = 0.5
out = image1 * (1.0 - alpha) + image2 *alpha
image[i+tileSize-1:i+tileSize+1,:] = out
for i in range(0,len(image[0])-tileSize,tileSize):
image1 = image[:,i+tileSize-2:i+tileSize]
image2 = image[:,i+tileSize:i+tileSize+2]
alpha = 0.5
out = image1 * (1.0 - alpha) + image2 *alpha
image[:,i+tileSize-1:i+tileSize+1]=out
b = 1
return image
#gausian bleninding of the individual pictures does not currently work, http://docs.opencv.org/3.1.0/dc/dff/tutorial_py_pyramids.html#gsc.tab=0
def blend(image, tilesize):
b=1
a = len(image)
l = len(image[0])
p= image[0:tileSize,30:tileSize]
for j in range(0,len(image)-tileSize,tileSize):
for k in range(0,len(image[0])-tileSize,tileSize):
A = image[j+tileSize-3:j+tileSize,k+tileSize-3:k+tileSize]
G = A.copy()
gpA = [G]
for i in xrange(3):
G = cv2.pyrDown(G)
gpA.append(G)
B = image[j:j+tileSize,k+tileSize:k+tileSize+tileSize]
G = B.copy()
gpB = [G]
for i in xrange(3):
G = cv2.pyrDown(G)
gpB.append(G)
lpA = [gpA[2]]
for i in xrange(2,0,-1):
GE = cv2.pyrUp(gpA[i])
b = GE[0:len(gpA[i-1]),0:len(gpA[i-1])]
L = cv2.subtract(gpA[i-1],b)
lpA.append(L)
lpB = [gpB[2]]
for i in xrange(2,0,-1):
GE = cv2.pyrUp(gpB[i])
b = GE[0:len(gpB[i-1]),0:len(gpB[i-1])]
L = cv2.subtract(gpB[i-1],b)
lpB.append(L)
# Now add left and right halves of images in each level
LS = []
for la,lb in zip(lpA,lpB):
rows,cols,dpt = la.shape
p =la[:,0:2]
pp= lb[:,cols/2:]
ls = np.hstack((la[:,0:2], lb[:,cols/2:]))
LS.append(ls)
ls_ = LS[0]
# now reconstruct
for i in xrange(1,3):
ls_ = cv2.pyrUp(ls_)
b = ls_[0:len(LS[i]),0:len(LS[i])]
ls_ = cv2.add(b, LS[i])
#cv2.imwrite('Pyramid_blending2.jpg',ls_)
image[j+tileSize-2:j+tileSize-2+tileSize,k+tileSize-2:k+tileSize-2+tileSize] = ls_
return image
if __name__ == "__main__":
#start with a tiles size of 30, this is almost always too big to work
tileSize = int(25)
#the mosaic function returns the mosaice image and the Mean Squared Error
mes, inputImage = mosaic(tileSize)
i = 0
print tileSize
#cv2.imshow("Progress", inputImage)
#cv2.waitKey(10000)
cv2.imwrite(str('pictures/out/original.jpg'), inputImage)
s = inputImage.size
s = s/3
# s=s/134000
s = 0.000005 * s
print s
#if image blending is turned on a good mes is under 5000, if it is turned off then under 9000 is good. if it does not reach that threshold it will
#subtract 4 from the tile size and try again, as long as the tile size is not less then 8.
while i == 0:
if mes < 5000 or tileSize < 8 or tileSize < s:
i = 1
cv2.imwrite(str('pictures/out/out.jpg'), inputImage)
print "the optimal tile size is "
print tileSize
print mes
else:
tileSize -=4
mes,inputImage = mosaic(tileSize)
s = inputImage.size
s = s/3
#s=s/134000
s = 0.000005 * s
print tileSize
print mes
print s