File size: 6,967 Bytes
80eac1a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
206
import os
import re
import numpy as np
import matplotlib.pyplot as plt
from matplotlib.widgets import MultiCursor
from scipy import ndimage
from matplotlib import cm
from PIL import Image, ImageDraw, ImageFont
from skimage import morphology
from skimage.measure import regionprops


def comp(*ims, figsize=(20, 10)):
    N = len(ims)
    ncols = {1: 1, 2: 2, 3: 3, 4: 2, 5: 3, 6: 3, 7: 4, 8: 4, 9: 3}
    nrows = {1: 1, 2: 1, 3: 1, 4: 2, 5: 2, 6: 2, 7: 2, 8: 2, 9: 3}
    fig, axes = plt.subplots(ncols=ncols[N], nrows=nrows[N], sharex=True, sharey=True, figsize=figsize)
    fig.subplots_adjust(wspace=0.01, hspace=0.01)
    axes = axes.ravel()
    cursor = MultiCursor(fig.canvas, axes,
                         horizOn=True, vertOn=True, color='red', linewidth=1)
    for i in range(N):
        axes[i].imshow(ims[i])
    return fig, axes, cursor


def load_im(fn):
    f = open(fn, 'rb')
    a = f.read()
    f.close()
    aa = str(a[:2048])
    xpix = int(re.findall('xpixels\s?=\s?([0-9]*)', aa)[0])
    ypix = int(re.findall('ypixels\s?=\s?([0-9]*)', aa)[0])
    a = a[2048:]

    words = [a[k * 2:k * 2 + 2] for k in range(xpix * ypix)]
    arr = [int.from_bytes(words[k], byteorder='little', signed=True) for k in range(len(words))]
    im = np.array(arr).reshape((ypix, xpix))

    im = (im.T - np.mean(im, axis=1) +
          np.mean(ndimage.gaussian_filter(im, 10), axis=1)).T  # palliate horizontal artifact
    im = im - np.min(im)
    im = im / np.max(im)  # normalize to 0.0-1.0
    return im


def G(x):
    return 1 - np.abs(x) ** .5


def growcut(land, labels, strength, maxiter=5):
    Ni, Nj = land.shape
    sidei, sidej = np.arange(Ni), np.arange(Nj)
    ij = np.dstack(np.meshgrid(sidei, sidej))[:, :, ::-1]
    iijj = np.tile(ij, (9, 1, 1, 1))
    for i, k in enumerate(((0, 0), (1, 0), (1, 1), (0, 1), (-1, 1), (-1, 0), (-1, -1), (0, -1), (1, -1))):
        iijj[i, :, :, :] += np.array(k)
        iijj[i, :, :, 0] = iijj[i, :, :, 0].clip(0, land.shape[0] - 1)
        iijj[i, :, :, 1] = iijj[i, :, :, 1].clip(0, land.shape[1] - 1)
    neigh_slice = np.s_[iijj[:, :, :, 0], iijj[:, :, :, 1]]

    this_labels = labels * 1
    this_strength = strength * 1

    neigh_val = land[neigh_slice]
    jump_diff = land - neigh_val
    g = G(jump_diff)

    for i in range(maxiter):
        # print(np.sum(this_labels), end=' ')
        neigh_lab = this_labels[neigh_slice] * 1
        neigh_str = this_strength[neigh_slice] * 1

        attack_force = g * neigh_str

        new_layer = np.argmax(attack_force, axis=0)
        new_lab = neigh_lab[new_layer, iijj[0, :, :, 0], iijj[0, :, :, 1]] * 1
        new_strength = attack_force[new_layer, iijj[0, :, :, 0], iijj[0, :, :, 1]] * 1

        this_labels = new_lab
        this_strength = new_strength

    return this_labels, this_strength


def pyramid_contrast(im):
    oom = []
    ms = []
    for d in (9, 15):  # (9, 11, 13, 15, 17,25):#(3, 6, 9, 12, 15, 18, 21):
        disk = morphology.disk(d)
        m = ndimage.percentile_filter(im, 10, footprint=disk)
        M = ndimage.percentile_filter(im, 90, footprint=disk)
        om = (im - m) / (M - m)
        om = np.nan_to_num(om).clip(0, 1)
        oom.append(om)
        ms.append(M - m)
    oom = np.array(oom)
    # ms = np.array(ms)
    land = np.mean(oom, axis=0)
    return land


def segmentate(land, alpha=0.7, beta=0.6):
    if alpha < beta:
        print("alpha must be greater than beta")
        assert False
    foreground = ndimage.binary_erosion(land > alpha, iterations=1)
    background = land < beta

    lab = ndimage.label(foreground)[0]
    lab[lab > 0] += 1
    lab[background] = 1

    strength = (lab > 1) * 1. + (lab == 1) * 1.
    this_labels, this_strength = growcut(land,
                                         lab, strength, maxiter=25)
    w = (this_labels != np.roll(this_labels, 1, axis=0)) + (this_labels != np.roll(this_labels, 1, axis=1))

    b = w * 0
    lab2 = ndimage.label(~w)[0]
    for l in np.unique(lab2)[1:]:
        if np.sum(foreground[l == lab2]) > 0:
            b[l == lab2] = 1

    lab2 = ndimage.label(ndimage.binary_dilation(this_labels > 1))[0]
    return lab2


def filter_objects(lab2, max_eccentricity=0.93, min_size=10, max_size=200, min_convex_coverage=0.8):
    props = regionprops(lab2)  # object metrics
    b = lab2 * 0.
    for i in np.unique(lab2)[1:]:
        ind = i - 1
        e = props[ind].eccentricity
        s = props[ind].area
        c = s * 1 / props[ind].convex_area
        # filter objects by eccentricity, size, ans convex hull coverage
        if e < max_eccentricity and (min_size < s < max_size) and c > min_convex_coverage:
            lev = 1
        else:
            lev = 2
        b[lab2 == i] = lev
    return b


def present(im, land, b):
    original_im = cm.afmhot(im)[:, :, :3]
    # Contrast Level 0.5
    resim = 0.5 * land + (1 - 0.5) * im
    enhanced_im = cm.afmhot(resim)[:, :, :3]
    monochrome_land = np.tile(land, (3, 1, 1)).transpose((1, 2, 0))
    detected = b == 1
    detected = ndimage.binary_dilation(detected)  # * (~ndimage.binary_erosion(detected))
    filtered = b == 2
    filtered = ndimage.binary_dilation(filtered)  # * (~ndimage.binary_erosion(filtered))

    monochrome_land[detected] *= np.array([.3, 1, .3])
    monochrome_land[filtered] *= np.array([1, .6, .6])

    newim = np.hstack((enhanced_im, monochrome_land))
    newim = np.dstack((newim, newim[:, :, 0] * 0 + 1))

    base = Image.fromarray((newim * 255).astype(np.uint8))
    original_im = Image.fromarray((original_im * 255).astype(np.uint8))
    enhanced_im = Image.fromarray((enhanced_im * 255).astype(np.uint8))

    # make a blank image for the text, initialized to transparent text color
    txt = Image.new("RGBA", base.size, (255, 255, 255, 0))

    # get a font
    # fnt = ImageFont.truetype("/usr/share/fonts/truetype/freefont/FreeMonoBold.ttf", 30, encoding="unic")
    fnt = ImageFont.load_default()
    # get a drawing context
    d = ImageDraw.Draw(txt)
    ct = np.max(ndimage.label(b == 1)[0]) - 1
    d.text((600, 10), "CNOs: {:d}".format(ct), font=fnt, fill=(255, 50, 50, 255))
    out = Image.alpha_composite(base, txt)

    return out, original_im, enhanced_im, ct


def treat_one_image(fn, growcut_path, original_png_path, enhanced_png_path):

    file_list = []
    growcut_list = []

    # load data
    im = load_im(fn)

    # pyramid contrast
    land = pyramid_contrast(im)

    # detect objects
    lab2 = segmentate(land, alpha=.75, beta=0.7)

    # visualize
    b = filter_objects(lab2, max_eccentricity=0.967, min_size=30, max_size=200, min_convex_coverage=0.5)
    growcut_im, original_im, enhanced_im, ct = present(im, land, b)

    file_name = os.path.split(fn)[1][0:-10]

    original_im.save(os.path.join(original_png_path, file_name) + '.png')
    enhanced_im.save(os.path.join(enhanced_png_path, file_name) + '.png')
    growcut_im.save(os.path.join(growcut_path, file_name) + '.png')

    return file_name, ct