-
Notifications
You must be signed in to change notification settings - Fork 0
/
CCL_RLE.py
141 lines (101 loc) · 4.82 KB
/
CCL_RLE.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
# ------------------ Connected Component analysis ------------------ #
# Import libraries
import numpy as np
import operator
def intersection(I,J):
""" This function takes as input two tuples of the format (start,end) and returns a
tuple (start,end) representing the intersection interval of pixels. """
a=max(I[0],J[0])
b=min(I[1],J[1])
if a<=b:
return (a,b)
else:
return None
def CCL_RLE(im):
""" This function takes as input a binary image, searches for all connected objects and returns
a binary image containg only the biggest connected objects : lesion and background. """
# Labels matrix : defines the different objects represented with the same label
labels = np.zeros((im.shape[0]+1,im.shape[1]+2))
# im1 : the image padded with zeros in order to calculate labels of edge pixels
im1=np.ones((im.shape[0]+1,im.shape[1]+2))
im1[1:,1:-1]=im
# --------- First pass : construct labels matrix --------- #
labels_num=0 # number of labels
# Loop over the image pixels and check if the associated mask contains labels
for i in range(1,im1.shape[0]) :
for j in range(1,im1.shape[1]-1) :
if im1[i,j] != 1 :
L=[] # list of labels
# Append labels to L in case a pixel of the mask is equal to the pixel in question
if im1[i,j-1]==im1[i,j]:
L.append(labels[i,j-1])
if im1[i-1,j-1]==im1[i,j]:
L.append(labels[i-1,j-1])
if im1[i-1,j]==im1[i,j]:
L.append(labels[i-1,j])
if im1[i-1,j+1]==im1[i,j]:
L.append(labels[i-1,j+1])
# In case there are no labels identified, we add a new label to L
if len(L)==0:
labels_num+=1
labels[i,j]=labels_num
else:
# In case we find multiple labels, we assign the minimum one to the pixel in question
labels[i,j]=min(L)
# ----------- second pass : Run Length Encoding ----------- #
# This step determines the list of tuples used to determine the overlap region in the next step.
# Each tuple defines the start and end of the object area.
# L_codes is the list of all the tuples
L_codes = []
for i in range(1,im1.shape[0]) :
# L will contain the first and the last elements of objects for each row of the image
L = []
# Looping over each row and column in the image
for j in range(1,im1.shape[1]-1) :
# Insertion of the first index when the object first meets
if (im1[i,j-1]==1) and (im1[i,j]==0):
first=j
# Insertion of the last index of the object
if (im1[i,j]==0) and (im1[i,j+1]==1):
L.append((first,j))
# Add the tuple in the list
L_codes.append(L)
# Overlap (Merge)
# This step is used to find the overlapping areas.
# Looping over each tuple
for i in range(len(L_codes)-1):
# Finding the overlapped area
for l1 in L_codes[i]:
for l2 in L_codes[i+1]:
# Keeping the minimum value between the two overlapped areas
if intersection(l1,l2)!= None:
min_label = min(labels[i,l1[0]],labels[i+1,l2[0]])
# Changing the labels of each part with the minimum label found
labels[i,l1[0]:l1[1]+1] = min_label
labels[i+1,l2[0]:l2[1]+1] = min_label
labels=labels[1:,1:-1]
# ------------ Keep the two biggest regions ------------ #
# Define a dictionary D that defines labels as keys and the number of associated pixels as values
D={}
for i in range(labels.shape[0]):
for j in range(labels.shape[1]):
if labels[i,j] in D.keys():
D[labels[i,j]]+=1
else :
D[labels[i,j]]=1
# Search for the two biggest regions (those with the two maximum values in dictionary D)
Max_labels=[] # list of the two labels with biggest regions
# First biggest region
var=max(D.items(), key=operator.itemgetter(1))[0]
Max_labels.append(var)
D[var]=0
# Second biggest region
var=max(D.items(), key=operator.itemgetter(1))[0]
Max_labels.append(var)
D[var]=0
# Loop over the image pixels, if the label pixel doesn't figure in Max_labels list assign it as background
for i in range(im.shape[0]):
for j in range(im.shape[1]):
if labels[i,j] not in Max_labels:
im[i,j]=1
return(im)