-
Notifications
You must be signed in to change notification settings - Fork 5
/
Copy pathsegment.py
130 lines (108 loc) · 3.49 KB
/
segment.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
# IMPLEMENTING MEAN SHIFT IMAGE SEGMENTATION IN PYTHON
# Author(s): Pranshu Gupta, Abhishek Jain
###################################################################################################
import Image
import numpy as np
import time as t
import sys
from gaussian_mean import gaussian_mean
###################################################################################################
bandwidth = None
Bin = 40
kertype = "flat"
if len(sys.argv) == 4:
filename = str(sys.argv[1])
bandwidth = int(sys.argv[2])
gaussian = int(sys.argv[3])
else:
print "Usage: python segment.py bandwidth do_gaussian"
exit()
if gaussian == 1:
kertype = "gauss"
m = 1
S = 5
threshold = 1.0
print "Loading the Image " + filename + ".jpg"
img = Image.open("img/" + filename + ".jpg")
img.load()
img = np.array(img)
seg_img = img
rows, cols, dim = img.shape
meandist = np.array([[1000.0 for r in xrange(cols)] for c in xrange(rows)])
labels = np.array([[-1 for r in xrange(cols)] for c in xrange(rows)])
print "Running the Mean Shift algorithm ..."
start = t.time()
means = []
for r in xrange(0,rows,Bin):
for c in xrange(0,cols,Bin):
seed = np.array([r,c,img[r][c][0],img[r][c][1],img[r][c][2]])
for n in xrange(15):
x = seed[0]
y = seed[1]
r1 = max(0,x-Bin)
r2 = min(r1+Bin*2, rows)
c1 = max(0,y-Bin)
c2 = min(c1+Bin*2, cols)
kernel = []
for i in xrange(r1,r2):
for j in xrange(c1,c2):
dc = np.linalg.norm(img[i][j] - seed[2:])
ds = (np.linalg.norm(np.array([i,j]) - seed[:2]))*m/S
D = np.linalg.norm([dc,ds])
if D < bandwidth:
kernel.append([i,j,img[i][j][0],img[i][j][1],img[i][j][2]])
kernel = np.array(kernel)
# print kernel
if gaussian == 0:
mean = np.mean(kernel,axis=0,dtype=np.int64)
elif gaussian == 1:
mean = gaussian_mean(kernel, seed, bandwidth)
# Get the shift
dc = np.linalg.norm(seed[2:] - mean[2:])
ds = (np.linalg.norm(seed[:2] - mean[:2]))*m/S
dsm = np.linalg.norm([dc,ds])
seed = mean
if dsm <= threshold:
# print "Mean " + str(len(means)+1) + " converges in: " + str(n) + " iterations"
break
means.append(seed)
end = t.time()
print "Time taken for mean shift: " + str((end - start)/60) + " min"
print "Grouping together the means that are closer than the bandwidth ..."
flags = [1 for me in means]
for i in xrange(len(means)):
if flags[i] == 1:
w = 1.0
j = i + 1
while j < len(means):
dc = np.linalg.norm(means[i][2:] - means[j][2:])
ds = (np.linalg.norm(means[i][:2] - means[j][:2]))*m/S
dsm = np.linalg.norm([dc,ds])
if dsm < bandwidth:
means[i] = means[i] + means[j]
w = w + 1.0
flags[j] = 0
j = j + 1
means[i] = means[i]/w
converged_means = []
for i in xrange(len(means)):
if flags[i] == 1:
converged_means.append(means[i])
converged_means = np.array(converged_means)
# print "Number of Seeds: " + str(len(means))
# print "Number of Means: " + str(len(converged_means))
print "Constructing the segmented image ..."
for i in xrange(rows):
for j in xrange(cols):
for c in xrange(len(converged_means)):
dc = np.linalg.norm(img[i][j] - converged_means[c][2:])
ds = (np.linalg.norm(np.array([i,j]) - converged_means[c][:2]))*m/S
D = np.linalg.norm([dc,ds])
if D < meandist[i][j]:
meandist[i][j] = D
labels[i][j] = c
seg_img[i][j] = converged_means[labels[i][j]][2:]
print "Saving the segmented image ..."
seg_img = Image.fromarray(seg_img)
seg_img.save("img/" + kertype + "_output_" + filename + "_" + str(bandwidth) + ".jpg")
print bandwidth, len(converged_means)