e2dhipseg/label.py at master · MICLab-Unicamp/e2dhipseg · GitHub

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
'''
Interface with ia870 toolbox (https://github.com/MICLab-Unicamp/ia870)
'''

import numpy as np
from PIL import Image
from ia870 import ialabel
from utils import show_grayimage, ESC
import cv2 as cv
import time


def get_largest_components(volume, mask_ths=0.5, ncomponents=2, debug=False):
    '''
    Returns 2 or 1 largest components, input should be numpy in cpu
    '''
    begin = time.time()

    volume = volume > mask_ths

    if volume.astype(np.int).sum() == 0:
        return volume.astype(np.float32)

    labels = ialabel(volume, np.ones((3, 3, 3), dtype=bool))

    bincount = np.bincount(labels.flat)
    if debug:
        print("label count: {}".format(bincount))

    # Increasing sort (0, 1, 2 ,3...)
    sorted_count = np.sort(bincount)

    # Gets two labels in bincount with more area, excluding the object with higher area (background)
    largest_components = {}

    upper_limit = 4
    if len(bincount) == 2:
        upper_limit = 3

    for i in range(2, upper_limit):
        label = np.where(bincount == sorted_count[-i])[0][0]
        largest_components[str(label)] = sorted_count[-i]

    if debug:
        print("two highest labels excluding highest {}".format(largest_components))
        print("largest components processing time: {}s".format(time.time() - begin))

    highest_labels = np.asarray(list(largest_components.keys()), dtype=np.int64)
    if debug:
        print("Largest object labels: {}".format(highest_labels))

    # print(labels.max())
    # print(labels.min())
    # print(labels.dtype)
    hlabel_len = len(highest_labels)
    if hlabel_len >= 2:
        return (np.ma.masked_not_equal(labels, highest_labels[0]).filled(0) + np.ma.masked_not_equal(labels,
                highest_labels[1]).filled(0)).astype(np.bool).astype(np.float32)
    else:
        return np.ma.masked_not_equal(labels, highest_labels[0]).filled(0).astype(np.bool).astype(np.float32)


def ialabel_test():
    '''
    Tests labeling in sample 2D and 3D numpy
    '''

    # 2D numpy array from 0 to 24
    f = np.arange(24).reshape(4, 6) > 14

    # Label the 2D image, works!
    ialabel(f)

    # Label barcode
    fbarcode = Image.open('data/barcode.tif')
    show_grayimage("Barcode image", fbarcode)

    # Threhsold to binary image
    fbarc = np.array(fbarcode) < 128
    show_grayimage("Threshold barcode image", fbarc)

    # Label components of binary barcode image
    flabel = ialabel(fbarc)
    print("How many components in 2D test: {}".format(flabel.max()))
    show_grayimage("Labeled 2D image", flabel.astype(np.uint8))

    h = np.bincount(flabel.flat)
    print("Count for each label {}".format(h))

    print("Biggest component index {}".format(np.argmax(h[1:])))

    show_grayimage("Biggest component", (flabel == 340).astype(np.uint8)*255)

    print("3D test running in random 181, 217, 181 shape")

    f3 = np.zeros((181, 217, 181))
    cubesize = 20
    ncubes = 180//cubesize - 1
    for i in range(0, ncubes, 2):  # 2 step to give some empty space
        cubeshape = (cubesize + i, cubesize + i, cubesize + i)
        print("Inserting cube {} with shape {}".format(i//2, cubeshape))
        cube = np.ones(cubeshape)
        j = i*cubesize
        f3[j:j+cubeshape[0], j:j+cubeshape[0], j:j+cubeshape[0]] = cube

    for i in range(217):
        cv.imshow("before", f3[:, i, :])
        cv.waitKey(10)

    f3 = get_largest_components(f3)
    print(f3.dtype)
    for i in range(217):
        cv.imshow("after", f3[:, i, :])
        cv.waitKey(10)

    quit()
    # plt.show()

    cap = cv.VideoCapture(0)
    cap.set(cv.CAP_PROP_FRAME_WIDTH, 800)
    cap.set(cv.CAP_PROP_FRAME_HEIGHT, 600)

    while True:
        _, img = cap.read()
        # gimg = cv.cvtColor(img, cv.COLOR_BGR2GRAY)
        gimg = img
        timg = gimg > 128
        labeledint = ialabel(timg, np.ones((3, 3, 3), dtype=bool))
        labeled = labeledint.astype(np.float32)
        print("nlabels: {}".format(labeled.max()))

        labeled = (labeled - labeled.min()) / (labeled.max() - labeled.min())
        gimg = (gimg - gimg.min()) / (gimg.max() - gimg.min())
        timg = timg.astype(np.float32)

        biggest_labels = np.bincount(labeledint.flat)
        print(biggest_labels)
        biggest_label = np.argmax(biggest_labels)

        cv.imshow("5 Biggest Components", cv.resize(np.hstack((gimg, timg, labeled,
                                                               (labeledint == biggest_label).astype(np.float32))), (1920, 600)))
        if cv.waitKey(1) == ESC:
            quit()


if __name__ == "__main__":
    ialabel_test()