-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathdct.py
157 lines (134 loc) · 5.96 KB
/
dct.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
import cv2
import numpy as np
import numpy.fft as fft
from PIL import Image, ImageDraw, ImageFont
import matplotlib.pyplot as plt
from matplotlib import rcParams
rcParams['font.family'] = 'monospace'
rcParams['font.monospace'] = ['Ubuntu Mono', 'JetBrains Mono', 'Consolas', 'monospace']
WATER_MARK_ALPHA = 0.04
TAMPER_LOCATE_CROP = 0.05
BLUR_SIZE = 128
block_size = (32,32,3)
# 文字转图片，字体大小填充图片高度
def text_to_img(text, img_height):
font = ImageFont.truetype('consola', img_height)
img_width = int(font.getlength(text))
img = Image.new('RGB', (img_width, img_height), (0, 0, 0))
draw = ImageDraw.Draw(img)
draw.text((0, 0), text, font=font, fill=(255, 255, 255))
return np.array(img)
# 载入适合频域的水印
# 如果是 type 为 text 则生成内容为 name 的图片作为水印
# 如果是 type 为 image 则载入目录下名称为 name 的图片作为水印, 将图片等比缩放至高度为 height
def load_water_mark(img:np.ndarray,type: str, name: str, height: int = None):
if type == 'text':
wm = text_to_img(name, height)
elif type == 'image':
wm = cv2.imread(name)
if height is not None:
wm = cv2.resize(img, (int(img.shape[1] * height / img.shape[0]), height))
else:
raise Exception('type must be text or image')
return wm
# 制作水印
# 创建 target 大小的图片
def make_water_mark(target: np.ndarray, water_mark: np.ndarray, offset: tuple = (0, 0)):
wm = np.zeros(target.shape, dtype=np.uint8)
# 制作
wm[-water_mark.shape[0]-offset[0]:-offset[0], -water_mark.shape[1]-offset[1]:-offset[1]] = water_mark
wm = wm[::-1, ::-1]
wm[-water_mark.shape[0]-offset[0]:-offset[0], -water_mark.shape[1]-offset[1]:-offset[1]] = water_mark
wm = wm[::-1, ::-1]
return np.uint8(wm)
# 图像显示辅助函数
# 并排显示 n*m 张图像，n行m列
# 在每张图左侧和顶侧显示像素坐标，在每张图顶侧显示标题
def show_images(images: list, titles: list, n: int, m: int):
plt.figure(dpi=800)
plt.subplots_adjust(wspace=0.5, hspace=0.5)
for i in range(n):
for j in range(m):
plt.subplot(n, m, i * n + j + 1)
plt.imshow(images[i * n + j], cmap='gray')
plt.title(titles[i * n + j])
plt.show()
# 频谱中心化、对数化、归一化、*255
def center_log_spectrum(spectrum: np.ndarray):
spectrum = np.log(np.abs(spectrum) + 1)
spectrum = (spectrum - np.min(spectrum))*255 // (np.max(spectrum) - np.min(spectrum))
return np.uint8(spectrum)
def rgb_fft2(rgbImage):
# 对rgb图像分通道做fft2，并将高频分量移至中心
if(rgbImage.__class__==np.ndarray):
return fft.fftshift(fft.fft2(rgbImage,axes=(0,1)),axes=(0,1))
elif(rgbImage.__class__==list):
return [fft.fftshift(fft.fft2(i,axes=(0,1)),axes=(0,1)) for i in rgbImage]
else:
return None
def rgb_ifft2(rgbImage_fft_shift):
# 上一函数的逆过程
if(rgbImage_fft_shift.__class__==np.ndarray):
return fft.ifft2(fft.ifftshift(rgbImage_fft_shift,axes=(0,1)),axes=(0,1))
elif(rgbImage_fft_shift.__class__==list):
return [fft.ifft2(fft.ifftshift(i,axes=(0,1)),axes=(0,1)) for i in rgbImage_fft_shift]
else:
return None
def rgb_dct2(rgbImage:np.ndarray):
'''
对rgbImage作dct变换，返回数据类型为float32的频域图像
'''
return np.transpose([cv2.dct(np.float32(rgbImage[:,:,i])) for i in range(3)],axes=(1,2,0))
def rgb_idct2(rgbImage_dct:np.ndarray):
'''
对rgbImage_idct作逆dct变换，返回数据类型为uint8的空域图像
'''
Img_idct2 = np.transpose([cv2.idct(rgbImage_dct[:,:,i]) for i in range(3)],axes=(1,2,0))
Img_idct2_qua = Img_idct2.copy()
for i in range(Img_idct2_qua.shape[0]):
for j in range(Img_idct2_qua.shape[1]):
for k in range(Img_idct2_qua.shape[2]):
if(Img_idct2_qua[i,j,k]>255):
Img_idct2_qua[i,j,k]=255
elif(Img_idct2_qua[i,j,k]<0):
Img_idct2_qua[i,j,k]=0
else:
Img_idct2_qua[i,j,k]=round(Img_idct2_qua[i,j,k])
Img_idct2_qua = np.uint8(Img_idct2_qua)
return Img_idct2_qua
def get_feature(Image:np.ndarray,block_size:tuple):
feature = np.zeros(shape=(Image.shape[0]//block_size[0],Image.shape[1]//block_size[1],3))
for i in range(feature.shape[0]):
for j in range(feature.shape[1]):
feature[i,j,:] = np.max( Image[block_size[0]*i : block_size[0]*(i+1),block_size[1]*j : block_size[1]*(j+1),:],axis=(0,1))
feature = np.uint8((feature - np.min(feature))*255 // (np.max(feature) - np.min(feature)))
return feature
def differ(img1_fea:np.ndarray,img2_fea:np.ndarray,return_pos=True):
sum = 0
position = []
for i in range(img1_fea.shape[0]):
for j in range(img1_fea.shape[1]):
if(np.max(np.abs(img1_fea[i,j,:]-img2_fea[i,j,:])) > 100):
sum = sum+1
if(return_pos == True):
position.append((i,j))
if(return_pos == True):
return position,sum/(img1_fea.shape[0]*img1_fea.shape[1])
else:
return sum/(img1_fea.shape[0]*img1_fea.shape[1])
def extract(Image:np.ndarray,wm_size:tuple):
wm_ext = np.zeros(shape=wm_size)
Image_dct = rgb_dct2(Image)
for i in range(wm_size[0]):
for j in range(wm_size[1]):
block = Image[i*block_size[0]:(i+1)*block_size[0],j*block_size[1]:(j+1)*block_size[1],:]
block_dct = rgb_dct2(block)
wm_ext[i,j,:] = block_dct[-1,-1,:]
return wm_ext
def mark_diff(Image,position,block_shape):
Image_marked = Image.copy()
for pos in position:
start_point = (pos[0]*block_shape[0],pos[1]*block_shape[1])
end_point = (pos[0] * block_shape[0] + block_shape[0] - 4, pos[1] * block_shape[1] + block_shape[1] - 4)
cv2.rectangle(Image_marked, start_point, end_point, (0, 0, 255), thickness=2, lineType=cv2.LINE_8)
return Image_marked