-
Notifications
You must be signed in to change notification settings - Fork 18
/
Copy pathtransform.py
171 lines (137 loc) · 5.17 KB
/
transform.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
import numpy as np
import math
import cv2
from corner_detection import detect_corners
def build_equ(four_corners):
X = np.zeros(shape=(4, 4))
for i, (x0, y0) in enumerate(four_corners):
X[i, :] = [x0, y0, x0 * y0, 1]
return X
def warpAffine(src_img, detected_corner,size):
'''
:param src_img:
:param detected_corner:
:param size: 目标大小
:return:
'''
height, width = size
target_corner = np.array([[0, 0], [0, width - 1], [height - 1, width - 1], [height - 1, 0]], dtype=np.int32)
# 计算x,y变化矩阵T_x,T_y
leftMatrix = build_equ(target_corner)
inversed_leftMat = np.linalg.inv(leftMatrix)
X1 = detected_corner[:, 0]
T_x = np.matmul(inversed_leftMat, X1)
Y1 = detected_corner[:, 1]
T_y = np.matmul(inversed_leftMat, Y1)
tar_img = fast_bi_inter(src_img, height, width, T_x, T_y)
return tar_img
def fast_bi_inter(src_img, height, width, T_x, T_y):
"""
使用矩阵计算，实现快速双线性插值
:param src_img: src image
:param height: target height
:param width: target width
:return:
"""
row_same = np.arange(0, height)
row_same = row_same[:, np.newaxis]
row_same = np.tile(row_same, (1, width))
col_same = np.arange(0, width)
col_same = col_same[np.newaxis, :]
col_same = np.tile(col_same, (height, 1))
mix_para = row_same * col_same
ones_para = np.ones(shape=(height, width))
paras = np.stack((row_same, col_same, mix_para, ones_para), axis=2)
paras = np.reshape(paras, (height * width, 4))
# 找到原图中的坐标
x = np.matmul(paras, T_x)
y = np.matmul(paras, T_y)
# 确定p1,p2,q1,q2
x_ceil = np.ceil(x).astype(int)
x_floor = np.floor(x).astype(int)
y_ceil = np.ceil(y).astype(int)
y_floor = np.floor(y).astype(int)
p1 = x - x_floor
# print(p1.shape) #(190512,)
p2 = 1 - p1
q1 = y - y_floor
q2 = 1 - q1
tar_img = np.expand_dims(p2 * q2, axis=1) * src_img[x_floor, y_floor, :] + \
np.expand_dims(p1 * q2, axis=1) * src_img[x_ceil, y_floor, :] + \
np.expand_dims(q1 * p1, axis=1) * src_img[x_ceil, y_ceil, :] + \
np.expand_dims(q1 * p2, axis=1) * src_img[x_floor, y_ceil, :]
tar_img = np.reshape(tar_img, (height, width, 3))
return tar_img
def main():
# 图片角点顺序如下：
# 0 1
# 3 2
# src_img_path="./data/000026.jpg"
# detected_corner = [[87,96], [99,330], [459,339], [440,57]]
src_img_path = "./data/000872.jpg"
detected_corner = [[119, 74], [129, 296], [442, 306], [444, 55]]
detected_corner = np.array(detected_corner, dtype=np.int32)
height, width = [504, 378]
target_corner = np.array([[0, 0], [0, width - 1], [height - 1, width - 1], [height - 1, 0]], dtype=np.int32)
# 计算x,y变化矩阵T_x,T_y
leftMatrix = build_equ(target_corner)
inversed_leftMat = np.linalg.inv(leftMatrix)
X1 = detected_corner[:, 0]
T_x = np.matmul(inversed_leftMat, X1)
Y1 = detected_corner[:, 1]
T_y = np.matmul(inversed_leftMat, Y1)
# test = np.matmul(leftMatrix[2, :], T_y)
# print(test)
# 双线性插值
src_img = cv2.imread(src_img_path)
# pick_img=cv2.copyMakeBorder(src_img,0,1,0,1,cv2.BORDER_REPLICATE)
tar_img = np.zeros(shape=(height, width, 3))
# 遍历每个像素，进行后向插值
for channel in range(3):
for i in range(height):
for j in range(width):
equation_coefficient = np.array([i, j, i * j, 1])
x = np.matmul(equation_coefficient, T_x)
y = np.matmul(equation_coefficient, T_y)
x_ceil = math.ceil(x)
x_floor = math.floor(x)
y_ceil = math.ceil(y)
y_floor = math.floor(y)
p1 = x - x_floor
p2 = 1 - p1
q1 = y - y_floor
q2 = 1 - q1
try:
tar_img[i, j, channel] = p2 * q2 * src_img[x_floor, y_floor, channel] + \
src_img[x_ceil, y_floor, channel] * p1 * q2 + \
src_img[x_ceil, y_ceil, channel] * q1 * p1 + \
src_img[x_floor, y_ceil, channel] * q1 * p2
except:
print(x_ceil, x_floor, y_ceil, y_floor)
print(src_img.shape)
assert 1 == 9
cv2.imshow('Window', np.uint8(tar_img))
cv2.waitKey(0)
def cvtColor(src_img):
"""
转彩色图像为灰度图像。模仿opencv命名
:param src_img: 彩色图像
:return: 灰度图像
"""
gray_img = np.sum(src_img, axis=2) / 3
return gray_img.astype(np.uint8)
if __name__ == "__main__":
# path = "./data/1.jpg"
path = "./data/000026.jpg"
# path = './data/000872.jpg'
# path = './data/001201.jpg'
# path = './data/001402.jpg'
# path = './data/001552.jpg'
src_img = cv2.imread(path)
gray_img = cvtColor(src_img)
detect_img, detected_corner = detect_corners(gray_img)
tar_img = warpAffine(src_img)
cv2.imshow('Window', np.uint8(tar_img))
cv2.waitKey(0)
cv2.imshow('dst', detect_img)
cv2.waitKey(0)