-
Notifications
You must be signed in to change notification settings - Fork 3
/
edgeExtractAllDemo.py
208 lines (178 loc) · 9.83 KB
/
edgeExtractAllDemo.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
import cv2 as cv
import numpy as np
from Algorithm.propagationLabel import getEdgesFromLabel, evaluateMeritForEdge
from skimage.measure import label
import matplotlib.pyplot as plt
from skimage import feature,morphology
import time
import os
def meanGray(gray, threshold):
lowSum = 0.0
lowNum = 0
highNum = 0
highSum = 0.0
for i in range(0, gray.shape[0]):
for j in range(0, gray.shape[1]):
if (gray[i, j] > threshold):
highSum += gray[i, j]
highNum = highNum + 1
elif (gray[i, j] <= threshold):
lowSum += gray[i, j]
lowNum = lowNum + 1
meanhigh = highSum*1.0 / highNum
meanlow = lowSum*1.0 / lowNum
L1 = (meanhigh + meanlow) / 2
return (meanhigh, meanlow, L1)
def iteractiveMethod(gray):
T_max = int(gray.max())
T_min = int(gray.min())
L0 = (T_max + T_min)*1.0 / 2
threshold = 0
while(1):
(meanhigh, meanlow, L1) = meanGray(gray, L0)
if(L1 == L0):
threshold = L0
break
L0 = L1
return threshold
def showHist(grayImage, figureName):
plt.figure(figureName)
hist = cv.calcHist([grayImage], [0], None, [256], [0, 256])
plt.subplot(121), plt.title(u"Original"), plt.imshow(grayImage, cmap="gray")
plt.subplot(122), plt.title(u"Grayscale histogram"), plt.plot(hist)
plt.show()
def denoiseByArea(grayImage, areaThresh, neighbors = 8):
"""
Function： Remove noises with area less than a certain threshold
Input： grayImage: grayscale image
areaThresh: threshold of area
neighbors: 4 Neighbors or 8 Neighbors (optional, default 4 Neighbors)
Output: grayImage without noises
"""
labeled, num = label(grayImage, neighbors=neighbors, return_num=True)
zerosMatrix = np.zeros(labeled.shape)
for index in range(1, num+1):
zerosMatrix[labeled == index] = 1
temp = np.count_nonzero(zerosMatrix)
if temp < areaThresh:
grayImage[labeled == index] = 0
zerosMatrix = np.zeros(labeled.shape)
return grayImage
if __name__ == "__main__":
# *************************************** Init ******************************************************
slices = ['001', '002', '003', '004']
numSlices = len(slices)
otsuF = np.zeros(numSlices)
iteractiveF = np.zeros(numSlices)
cannyF = np.zeros(numSlices)
meanF = np.zeros(numSlices)
gaussianF = np.zeros(numSlices)
F_average = np.zeros(5)
addressProject = os.path.join(os.getcwd(), "images") # Image directory
nextOriginalImageAddress = addressProject + "\\Original\\" # Original images directory
nextHumanLabeledRGBAddress = addressProject + "\\GroundTruth\\" # Ground truth images directory
resultsAddress = addressProject + "\\Results\\Others\\" # Storage directory of segmentation results
for index in range(0, numSlices): # A total of 5 series of images
nextOriginalImage = cv.cvtColor(cv.imread(nextOriginalImageAddress+slices[index]+".tif"), cv.COLOR_BGR2GRAY)
nextHumanLabeledRGB = cv.imread(nextHumanLabeledRGBAddress+slices[index]+".tif")
rowNumber = nextHumanLabeledRGB.shape[0]
colNumber = nextHumanLabeledRGB.shape[1]
nextHumanLabeledGray = np.zeros((rowNumber, colNumber))
nextHumanLabeledGray = cv.cvtColor(nextHumanLabeledRGB, cv.COLOR_BGR2GRAY)
(nextHumanLabeled, nNumber) = label(nextHumanLabeledGray, neighbors=4, return_num=True)
nextHumanLabeled = nextHumanLabeled.astype(np.int32)
nextGroundTruthEdge = np.zeros((rowNumber, colNumber))
nextGroundTruthEdge = getEdgesFromLabel(nextHumanLabeled)
nextGroundTruthEdge_fc = morphology.skeletonize(nextGroundTruthEdge / 255) * 255
cv.imwrite(resultsAddress+slices[index]+"-gt.tif", nextGroundTruthEdge_fc)
print("segment image ："+slices[index])
start_time = time.time()
# ********************************* Otsu method *****************************************
threshold, imgOtsu = cv.threshold(nextOriginalImage, 0, 255, cv.THRESH_BINARY_INV + cv.THRESH_OTSU)
# ********************************* Iterative threshold method ****************************************
threshold_iM = iteractiveMethod(nextOriginalImage)
T_two, imgIteractiveSeg = cv.threshold(nextOriginalImage, threshold_iM, 255, cv.THRESH_BINARY_INV) # 阈值化处理，阈值为：155
# ********************************* Sobel method ****************************************
# x = cv.Sobel(nextOriginalImage, cv.CV_16S, 1, 0)
# y = cv.Sobel(nextOriginalImage, cv.CV_16S, 0, 1)
# absX = cv.convertScaleAbs(x) # turn to uint8
# absY = cv.convertScaleAbs(y)
# imgSobel = cv.addWeighted(absX, 0.5, absY, 0.5, 0)
#
# ******************************* Adaptive threshold method **************************************
blurredGaussian = cv.GaussianBlur(nextOriginalImage, (3, 3), 0) # Gaussian filter
imgThreshMean = cv.adaptiveThreshold(blurredGaussian, 255, cv.ADAPTIVE_THRESH_MEAN_C, cv.THRESH_BINARY_INV, 5, 4)
# imgThreshMean = denoiseByArea(imgThreshMean, 300, neighbors=4)
# imgThreshMean = morphology.skeletonize(imgThreshMean / 255) * 255
# cv2.ADAPTIVE_THRESH_GAUSSIAN_C：Calculate the neighborhood weighted average as a threshold
imgThreshGaussian = cv.adaptiveThreshold(blurredGaussian, 255, cv.ADAPTIVE_THRESH_GAUSSIAN_C, cv.THRESH_BINARY_INV, 5, 4)
# imgThreshGaussian = denoiseByArea(imgThreshGaussian, 300, neighbors=4)
# imgThreshGaussian = morphology.skeletonize(imgThreshGaussian / 255) * 255
# ********************************** Canny method *****************************************
imgCanny = cv.Canny(blurredGaussian, 150, 200)
# ********************************** Dilation and erosion *****************************************
# kernel = cv.getStructuringElement(cv.MORPH_CROSS, (5, 5))
# imgEroded = cv.erode(nextOriginalImage, kernel) # erosion
# imgDilated = cv.dilate(nextOriginalImage, kernel) # Dilation
# imgClosed = cv.morphologyEx(nextOriginalImage, cv.MORPH_CLOSE, kernel) # close operation
# imgOpened = cv.morphologyEx(nextOriginalImage, cv.MORPH_OPEN, kernel) # open operation
#
# edgeSegI1 = cv.subtract(imgDilated, nextOriginalImage)
# edgeSegI2 = cv.subtract(nextOriginalImage, imgEroded)
# edgeSegI3 = cv.subtract(imgDilated, imgEroded)
# edgeSegI4 = cv.subtract(nextOriginalImage, imgOpened)
# edgeSegI5 = cv.subtract(imgClosed, nextOriginalImage)
# threshold, edgeSegI5 = cv.threshold(edgeSegI5, 0, 255, cv.THRESH_BINARY + cv.THRESH_OTSU)
# edgeSegI6 = cv.subtract(imgClosed, imgOpened)
# # showHist(edgeSegI6, "edgeSegI6")
# #threshold, edgeSegI6 = cv.threshold(edgeSegI6, 0, 255, cv.THRESH_BINARY + cv.THRESH_OTSU)
#
# edgeSegI7 = cv.subtract(imgDilated, imgClosed)
# edgeSegI8 = cv.subtract(imgOpened, imgEroded)
# edgeSegI9 = edgeSegI1.copy()
# for i in range(edgeSegI8.shape[0]):
# for j in range(edgeSegI8.shape[1]):
# if edgeSegI2[i, j] < edgeSegI1[i, j]:
# edgeSegI9[i, j] = edgeSegI2[i, j]
end_time = time.time()
print("it cost ", str(end_time-start_time), "sec")
# # # ********************************* evaluate ****************************************
print(u' figure of merit：')
# *************************Otsu*******************************
imgOtsu_fc = morphology.skeletonize(imgOtsu / 255) * 255
otsuF[index] = evaluateMeritForEdge(imgOtsu_fc, nextGroundTruthEdge_fc)
cv.imwrite(resultsAddress+slices[index]+"-Otsu.tif", imgOtsu_fc)
print(' Otsu F='+str(otsuF[index]))
F_average[0] += otsuF[index]
#
# ************************Iteractive****************************
imgIteractiveSeg_fc = morphology.skeletonize(imgIteractiveSeg / 255) * 255
iteractiveF[index] = evaluateMeritForEdge(imgIteractiveSeg_fc, nextGroundTruthEdge_fc)
cv.imwrite(resultsAddress+slices[index]+"-Iteractive.tif", imgIteractiveSeg_fc)
print(' Iteractive F='+str(iteractiveF[index]))
F_average[1] += iteractiveF[index]
#
# ***********************Canny*****************************************
imgCanny_fc = morphology.skeletonize(imgCanny / 255) * 255
cannyF[index] = evaluateMeritForEdge(imgCanny_fc, nextGroundTruthEdge_fc)
cv.imwrite(resultsAddress+slices[index]+"-Canny.tif", imgCanny_fc)
print(' Canny F='+str(cannyF[index]))
F_average[2] += cannyF[index]
#
# *************************ThreshMean*********************************
imgThreshMean_fc = morphology.skeletonize(imgThreshMean / 255) * 255
meanF[index] = evaluateMeritForEdge(imgThreshMean_fc, nextGroundTruthEdge_fc)
cv.imwrite(resultsAddress+slices[index]+"-adpativeMean.tif", imgThreshMean_fc)
print(' adpativeMean F='+str(meanF[index]))
F_average[3] += meanF[index]
#
# ************************** ThreshGaussian ********************************
imgThreshGaussian_fc = morphology.skeletonize(imgThreshGaussian / 255) * 255
gaussianF[index] = evaluateMeritForEdge(imgThreshGaussian_fc, nextGroundTruthEdge_fc)
cv.imwrite(resultsAddress+slices[index]+"-gaussian.tif", imgThreshGaussian_fc)
print(' gaussian F='+str(gaussianF[index]))
F_average[4] += gaussianF[index]
for av in range(0, 5):
F_average[av] = F_average[av]/len(slices)
print(' Otsu avF=', str(F_average[0]), ' Iteractive avF=', str(F_average[1]), ' Canny avF=', str(F_average[2]), ' adpativeMean avF=', str(F_average[3]), ' gaussian avF=', str(F_average[4]))
print("over")