CoinScanner/colorfeaturedetection.py at master · TjazVracko/CoinScanner · 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
import cv2
import numpy as np
from util import COIN_IMG_SIZE
from colormath import color_objects, color_diff


class ColorFeatureDetector:

    r = COIN_IMG_SIZE / 2  # NEW_SIZE / 2
    ym, xm = np.ogrid[-r:r, -r:r]
    coin_mask_1 = xm**2 + ym**2 > r**2  # ta maska definira krog (oziroma elemente zunaj kroga (manj nek rob) na kvadratu, saj teh ne upoštevamo)
    coin_mask_3 = np.dstack((coin_mask_1, coin_mask_1, coin_mask_1))

    # posebne maske za 1€ in 2€
    edge_width = COIN_IMG_SIZE / 8  # 25 pri NEW_SIZE=200
    coin_edge_mask_1 = (xm**2 + ym**2 > r**2) | (xm**2 + ym**2 < (r - edge_width)**2)
    coin_inside_mask_1 = xm**2 + ym**2 > (r - edge_width)**2
    coin_edge_mask_3 = np.dstack((coin_edge_mask_1, coin_edge_mask_1, coin_edge_mask_1))
    coin_inside_mask_3 = np.dstack((coin_inside_mask_1, coin_inside_mask_1, coin_inside_mask_1))

    @staticmethod
    def get_color_characteristics(coin_image):
        '''
        Retuns avarage color and standard deviation within coin cirlce
        glede na https://en.wikipedia.org/wiki/Color_difference
        se nam splača samo Lab prostor, in pol uporabimo tiste formule za razdaljo, ki kao predstavlja človeško zaznavo
        paper: http://www2.ece.rochester.edu/~gsharma/ciede2000/
        '''

        # spremenimo v float 0 do 1 image, da nam konverzija pol vrne prave vrednosti
        # fpi = coin_image.astype('float32') / 255
        fpi = np.float32(coin_image)
        fpi = fpi * (1.0/255)

        lab_coin = cv2.cvtColor(fpi, cv2.COLOR_BGR2Lab)
        masked_lab_coin = np.ma.array(lab_coin, mask=coin_mask_3)
        # avg_lab = masked_lab_coin.mean(axis=(0, 1))
        std_lab = masked_lab_coin.std(axis=(0, 1))

        masked_edge = np.ma.array(lab_coin, mask=coin_edge_mask_3)
        avg_edge = masked_edge.mean(axis=(0, 1))
        # std_edge = masked_edge.std(axis=(0, 1))

        masked_inside = np.ma.array(lab_coin, mask=coin_inside_mask_3)
        avg_inside = masked_inside.mean(axis=(0, 1))
        # std_inside = masked_inside.std(axis=(0, 1))

        return avg_edge.data, avg_inside.data, std_lab.data  # , std_inside.data  # (avg_gray, *avg_hsv, std_gray, *std_hsv)  # * je "Unpack" operator, RGB data se mi zdi neuporabna

    @staticmethod
    def get_color_histograms(coin_image):
        '''
        Returns nomalized histograms for edge and inside the coin
        for a and b channels seperately (4 hists total) of the CIE Lab space

        L←L∗255/100,a←a+128,b←b+128
        '''
        lab_coin = cv2.cvtColor(coin_image, cv2.COLOR_BGR2Lab)
        l, a, b = np.dsplit(lab_coin, 3)
        masked_a_edge = np.ma.array(a, mask=ColorFeatureDetector.coin_edge_mask_1)
        masked_b_edge = np.ma.array(a, mask=ColorFeatureDetector.coin_inside_mask_1)

        masked_a_inside = np.ma.array(b, mask=ColorFeatureDetector.coin_edge_mask_1)
        masked_b_inside = np.ma.array(b, mask=ColorFeatureDetector.coin_inside_mask_1)

        bins = np.arange(257)

        hist_a_edge, bins = np.histogram(masked_a_edge.compressed(), bins=bins, density=True)
        hist_b_edge, bins = np.histogram(masked_b_edge.compressed(), bins=bins, density=True)
        hist_a_inside, bins = np.histogram(masked_a_inside.compressed(), bins=bins, density=True)
        hist_b_inside, bins = np.histogram(masked_b_inside.compressed(), bins=bins, density=True)

        return hist_a_edge, hist_b_edge, hist_a_inside, hist_b_inside

    # https://docs.opencv.org/3.3.1/dd/d0d/tutorial_py_2d_histogram.html
    @staticmethod
    def get_2d_color_histograms_lab(coin_image):
        '''
        Retuns normalized 2d histogram of the a and b channels of L*a*b*
        '''
        lab_coin = cv2.cvtColor(coin_image, cv2.COLOR_BGR2Lab)
        l, a, b = np.dsplit(lab_coin, 3)

        masked_a_edge = np.ma.array(a, mask=ColorFeatureDetector.coin_edge_mask_1)
        masked_b_edge = np.ma.array(b, mask=ColorFeatureDetector.coin_edge_mask_1)

        masked_a_inside = np.ma.array(a, mask=ColorFeatureDetector.coin_inside_mask_1)
        masked_b_inside = np.ma.array(b, mask=ColorFeatureDetector.coin_inside_mask_1)

        bins = np.arange(257)
        # print(masked_a_edge.ravel().compressed().shape, "   ", masked_a_edge.ravel().compressed().shape)

        hist_edge, xbins, ybins = np.histogram2d(masked_a_edge.ravel().compressed(), masked_b_edge.ravel().compressed(), bins=bins, normed=True)
        hist_inside, xbins, ybins = np.histogram2d(masked_a_inside.ravel().compressed(), masked_b_inside.ravel().compressed(), bins=bins, normed=True)

        return hist_edge, hist_inside

    @staticmethod
    def get_2d_color_histograms_hsv(coin_image):
        '''
        Retuns normalized 2d histogram of the h and s channels of HSV
        '''
        hsv_coin = cv2.cvtColor(coin_image, cv2.COLOR_BGR2HSV)
        h, s, v = np.dsplit(hsv_coin, 3)

        masked_h_edge = np.ma.array(h, mask=ColorFeatureDetector.coin_edge_mask_1)
        masked_s_edge = np.ma.array(s, mask=ColorFeatureDetector.coin_edge_mask_1)

        masked_h_inside = np.ma.array(h, mask=ColorFeatureDetector.coin_inside_mask_1)
        masked_s_inside = np.ma.array(s, mask=ColorFeatureDetector.coin_inside_mask_1)

        bins = np.arange(257)
        # print(masked_a_edge.ravel().compressed().shape, "   ", masked_a_edge.ravel().compressed().shape)

        hist_edge, xbins, ybins = np.histogram2d(masked_h_edge.ravel().compressed(), masked_s_edge.ravel().compressed(), bins=bins, normed=True)
        hist_inside, xbins, ybins = np.histogram2d(masked_h_inside.ravel().compressed(), masked_s_inside.ravel().compressed(), bins=bins, normed=True)

        return hist_edge, hist_inside

    # https://python-colormath.readthedocs.io/en/latest/delta_e.html
    @staticmethod
    def color_difference(color1, color2):
        c1 = color_objects.LabColor(*color1)
        c2 = color_objects.LabColor(*color2)
        return color_diff.delta_e_cie2000(c1, c2)  # najnovejša formula, kao d best

    @staticmethod
    def color_difference_no_luminance(color1, color2):
        l = 50  # pač fiksni L
        c1 = color_objects.LabColor(l, color1[1], color1[2])
        c2 = color_objects.LabColor(l, color2[1], color2[2])
        return color_diff.delta_e_cie2000(c1, c2)  # najnovejša formula, kao d best

    @staticmethod
    def histogram_intersection(h1, h2, bins):
        bins = np.diff(bins)
        sm = 0
        for i in range(len(bins)):
            sm += min(bins[i]*h1[i], bins[i]*h2[i])
        return sm

    @staticmethod
    def histogram_2d_compare(h1, h2):
        return cv2.compareHist(np.array(h1, dtype='float32'), np.array(h2, dtype='float32'), cv2.HISTCMP_INTERSECT)