-
Notifications
You must be signed in to change notification settings - Fork 0
/
dwt.py
182 lines (142 loc) · 5.68 KB
/
dwt.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
#!usr/env/bin python3
# coding:utf-8
import cv2
import math
import pywt
import pywt.data
import numpy as np
import matplotlib.pyplot as plt
IMG_HEIGHT = 256
IMG_WIDTH = 256
MARK_HEIGHT = 128
MARK_WIDTH = 128
MARK_HEIGHT_HALF = int(MARK_HEIGHT / 2)
MARK_WIDTH_HALF = int(MARK_WIDTH / 2)
OFFSET_w1 = int((IMG_HEIGHT - MARK_HEIGHT) / 2) - 2
OFFSET_w2 = int((IMG_HEIGHT - MARK_HEIGHT) / 2) - 1
def split_watermark(watermark):
mark = watermark.copy()
mark = mark.reshape(-1, MARK_HEIGHT_HALF, MARK_WIDTH)
return mark[0], mark[1]
def insert(origin, watermark, offset_h, offset_w):
img = origin.copy()
mark = watermark.copy()
# binarization of 0 or 1
for r_idx, row in enumerate(mark):
for c_idx, col in enumerate(row):
mark[r_idx, c_idx] = 1 if col > 0 else 0
for row in range(0, MARK_HEIGHT_HALF):
for col in range(0, MARK_WIDTH):
tmp = img[row + offset_h, col + offset_w] + mark[row, col]
tmp = 255 if tmp > 255 else tmp
img[row + offset_h, col + offset_w] = tmp
return img
def extract(new, origin, offset_h, offset_w):
mark = new - origin
mark = mark[offset_h:offset_h + MARK_HEIGHT_HALF, offset_w:offset_w + MARK_WIDTH]
for r_idx, row in enumerate(mark):
for c_idx, col in enumerate(row):
mark[r_idx, c_idx] = 255 if col > 0 else 0
return mark
def retrieval(origin, embed_w1, embed_w2):
# wavelet transform
LL, (LH, HL, HH) = pywt.dwt2(embed_w1, 'haar')
_LL, (_LH, _HL, _HH) = pywt.dwt2(embed_w2, 'haar')
# extract part 1 of watermark
extract_w1 = extract(embed_w2, origin, OFFSET_w1, OFFSET_w1)
# extract part 2 of watermark
extract_w2 = extract(_HH, HH, OFFSET_w2, 0)
extract_w = np.concatenate((extract_w1, extract_w2))
return extract_w1, extract_w2, extract_w
# https://en.wikipedia.org/wiki/Peak_signal-to-noise_ratio
def psnr(img1, img2):
mse = np.mean((img1 / 1.0 - img2 / 1.0) ** 2)
if mse < 1.0e-10:
return 100
return 10 * math.log10(255.0 ** 2 / mse)
def cut_rectangle(image, h, w):
img = image.copy()
offset_h = int((IMG_HEIGHT - h) / 2)
offset_w = int((IMG_WIDTH - w) / 2)
for row in range(h):
for col in range(w):
img[(row + offset_h), (col + offset_w)] = None
return img
def cut_box_shape(image):
img = image.copy()
lw = 16 # line width
lw_half = int(lw / 2)
offset_h = int(IMG_HEIGHT / 4)
offset_w = int(IMG_WIDTH / 4)
h1, w1 = IMG_HEIGHT - offset_h * 2, IMG_WIDTH - offset_w * 2
h2, w2 = int(h1 / 2), int(w1 / 2)
h2_half, w2_half = int(h2 / 2), int(w2 / 2)
layer1 = np.zeros((h1, w1))
layer2 = np.full((h1 - lw, w1 - lw), 255)
layer3 = np.zeros((h2, w2))
layer4 = np.full((h2 - lw, w2 - lw), 255)
layer1[lw_half:h1 - lw_half, lw_half:w1 - lw_half] = layer2
layer3[lw_half:h2 - lw_half, lw_half:w2 - lw_half] = layer4
layer1[h2_half:h1 - h2_half, w2_half:w1 - w2_half] = layer3
mask = np.full((IMG_HEIGHT, IMG_WIDTH), 255)
mask[offset_h:IMG_HEIGHT - offset_h, offset_w:IMG_WIDTH - offset_w] = layer1
for r_idx, row in enumerate(mask):
for c_idx, col in enumerate(row):
img[r_idx, c_idx] = None if col == 0 else img[r_idx, c_idx]
return img
# Load image
origin = cv2.imread('origin.png', 0)
watermark = cv2.imread('watermark.png', 0)
_, watermark = cv2.threshold(watermark, 128, 255, cv2.THRESH_OTSU)
w1, w2 = split_watermark(watermark)
img_embed_w1 = insert(origin, w1, OFFSET_w1, OFFSET_w1)
# wavelet transform
LL, (LH, HL, HH) = pywt.dwt2(img_embed_w1, 'haar')
img_embed_HH = insert(HH, w2, OFFSET_w2, 0)
img_embed_w2 = LL, (LH, HL, img_embed_HH)
# inverse wavelet transform
img_embed_w2 = pywt.idwt2(img_embed_w2, 'haar')
print("Image PSNR:", '%.2f' % psnr(img_embed_w2, origin), "dB")
# embed & extract
extract_w1, extract_w2, extract_w = retrieval(origin, img_embed_w1, img_embed_w2)
fig = plt.figure(figsize=(12, 3))
ax = fig.add_subplot(1, 4, 1)
ax.imshow(img_embed_w1, interpolation="nearest", cmap=plt.cm.gray)
ax = fig.add_subplot(1, 4, 2)
ax.imshow(img_embed_w2, interpolation="nearest", cmap=plt.cm.gray)
ax = fig.add_subplot(2, 4, 3)
ax.imshow(extract_w1, interpolation="nearest", cmap=plt.cm.gray)
ax = fig.add_subplot(2, 4, 7)
ax.imshow(extract_w2, interpolation="nearest", cmap=plt.cm.gray)
ax = fig.add_subplot(1, 4, 4)
ax.imshow(extract_w, interpolation="nearest", cmap=plt.cm.gray)
fig.tight_layout()
plt.savefig('embed_and_extract.png')
# remove 128x128 region from image
cut_128x128 = cut_rectangle(img_embed_w2, 128, 128)
extract_w1, extract_w2, extract_w = retrieval(origin, img_embed_w1, cut_128x128)
fig = plt.figure(figsize=(12, 3))
ax = fig.add_subplot(1, 3, 1)
ax.imshow(cut_128x128, interpolation="nearest", cmap=plt.cm.gray)
ax = fig.add_subplot(2, 3, 2)
ax.imshow(extract_w1, interpolation="nearest", cmap=plt.cm.gray)
ax = fig.add_subplot(2, 3, 5)
ax.imshow(extract_w2, interpolation="nearest", cmap=plt.cm.gray)
ax = fig.add_subplot(1, 3, 3)
ax.imshow(extract_w, interpolation="nearest", cmap=plt.cm.gray)
fig.tight_layout()
plt.savefig('remove_128x128.png')
# remove box shape from image
cut_bs = cut_box_shape(img_embed_w2)
extract_w1, extract_w2, extract_w = retrieval(origin, img_embed_w1, cut_bs)
fig = plt.figure(figsize=(12, 3))
ax = fig.add_subplot(1, 3, 1)
ax.imshow(cut_bs, interpolation="nearest", cmap=plt.cm.gray)
ax = fig.add_subplot(2, 3, 2)
ax.imshow(extract_w1, interpolation="nearest", cmap=plt.cm.gray)
ax = fig.add_subplot(2, 3, 5)
ax.imshow(extract_w2, interpolation="nearest", cmap=plt.cm.gray)
ax = fig.add_subplot(1, 3, 3)
ax.imshow(extract_w, interpolation="nearest", cmap=plt.cm.gray)
fig.tight_layout()
plt.savefig('remove_box_shape.png')