-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathimg_clip.py
175 lines (166 loc) · 9.03 KB
/
img_clip.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
import mmcv
import os
import os.path as osp
import glob
import numpy as np
import math
from typing import List, Optional
from PIL import Image
from slice import *
def imageto8bit(image_path, save_8bit_path=None):
image, image_info, _ = read_multi_bands(image_path)
data_8bit = image.trans_img_16bit_to_8bit()
if save_8bit_path:
write_img(save_8bit_path, data_8bit.astype(np.uint8), img_geotrans=image_info[3], img_proj=image_info[4])
print("8位影像保存完成")
return data_8bit, image_info
def slice_index(img_data: np.ndarray,
stride: Optional[int] = None,
slice_size: int = 1024) -> List:
if len(img_data.shape) == 3:
single_band_size = img_data.shape[:2]
else:
single_band_size = img_data.shape
row_num = math.ceil(single_band_size[0] / slice_size) # 向上取整
col_num = math.ceil(single_band_size[1] / slice_size) # 向上取整
print(f"行列数：{single_band_size}，行分割数量：{row_num}，列分割数量：{col_num}")
slice_index = []
for i in range(row_num):
for j in range(col_num):
row_min = i * slice_size
row_max = (i + 1) * slice_size
if (i + 1) * slice_size > single_band_size[0]:
row_max = single_band_size[0]
col_min = j * slice_size
col_max = (j + 1) * slice_size
if (j + 1) * slice_size > single_band_size[1]:
col_max = single_band_size[1]
slice_index.append([row_min, row_max, col_min, col_max])
return slice_index
def slice(img_data: np.ndarray,
index: List,
slice_size: int = 1024,
edge_fill: bool = False,
fliter: bool = False,
percent: float = -1,
percent_nodata: float = 0) -> np.ndarray:
"""
"""
if fliter:
assert (percent <= 1) and (percent >= 0), 'percent 必须在0-1之间'
if percent_nodata:
assert (percent_nodata <= 1) and (percent_nodata >= 0), 'percent_nodata 必须在0-1之间'
rows = index[1] - index[0]
cols = index[3] - index[2]
if len(img_data.shape) > 2:
if edge_fill:
if fliter:
if (rows != slice_size) or (cols != slice_size):
if ((rows * cols / (slice_size * slice_size)) >= percent):
result = np.empty(shape=(slice_size, slice_size, img_data.shape[2]))
new_row_min = index[0] % slice_size
new_row_max = new_row_min + (index[1] - index[0])
new_col_min = index[2] % slice_size
new_col_max = new_col_min + (index[3] - index[2])
result[new_row_min:new_row_max, new_col_min:new_col_max, :] = img_data[index[0]:index[1],
index[2]:index[3], :]
else:
return None
else:
result = img_data[index[0]:index[1], index[2]:index[3], :]
else:
result = img_data[index[0]:index[1], index[2]:index[3], :]
else:
if edge_fill:
if fliter:
if (rows != slice_size) or (cols != slice_size):
if ((rows * cols / (slice_size * slice_size)) >= percent):
result = np.empty(shape=(slice_size, slice_size))
new_row_min = index[0] % slice_size
new_row_max = new_row_min + (index[1] - index[0])
new_col_min = index[2] % slice_size
new_col_max = new_col_min + (index[3] - index[2])
result[new_row_min:new_row_max, new_col_min:new_col_max] = img_data[index[0]:index[1],
index[2]:index[3]]
else:
return None
else:
result = img_data[index[0]:index[1], index[2]:index[3]]
else:
result = img_data[index[0]:index[1], index[2]:index[3]]
if len(result.shape) > 2:
if np.nansum(result) >= (percent_nodata * slice_size * slice_size * result.shape[2]):
return result.astype(img_data.dtype)
else:
return None
else:
if np.nansum(result) >= (percent_nodata * slice_size * slice_size):
return result.astype(img_data.dtype)
else:
return None
if __name__ == '__main__':
root = r'D:\LYH\dataset\greenhouse'
out_root = r'D:\LYH\dataset\greenhouse\samples_512'
save_8bit_dir = r'D:\LYH\dataset\water\0807\8bit'
slice_size = 512
transfor_8bit = False
prefixs = ['images', 'labels']
img_list = glob.glob(osp.join(root, prefixs[0]) + '/*.tif')
label_root = osp.join(root, prefixs[1])
# img_list = [r'D:\LYH\dataset\water\3612\GF1B_3612.tif']
# print(f"共读取影像{len(img_list)}景，输出路径为{out_root}")
# for img in img_list:
# img_data = mmcv.imread(img, backend='tifffile')
# # label_data = mmcv.imread(img.replace('tif', 'png'))
# label_data = Image.open(img.replace('.tif', '_label.tif'))
# label_data = np.asarray(label_data)
# print(label_data.shape)
# assert img_data.shape[0] == label_data.shape[0], '影像行数量与标注行数量不匹配'
# assert img_data.shape[1] == label_data.shape[1], '影像列数量与标注列数量不匹配'
# indexs = slice_index(img_data, slice_size=slice_size)
# percent = 0.2
# for i, idx in enumerate(indexs):
# img_slice_data = slice(img_data, idx, slice_size=slice_size, edge_fill=True, fliter=True, percent=percent)
# label_slice_data = slice(label_data, idx, slice_size=slice_size, edge_fill=True, fliter=True, percent=percent)
# if img_slice_data is not None:
# mmcv.imwrite(img_slice_data, osp.join(out_root, 'images', osp.basename(img).replace('.tif', '_' + str(i) + '.tif')))
# if label_slice_data is not None:
# # print(label_slice_data.shape)
# label_slice_data = Image.fromarray(np.uint8(label_slice_data > 0), mode='L')
# label_slice_data.save(osp.join(out_root, 'label', osp.basename(img).replace('.tif', '_' + str(i) + '.tif')))
# # mmcv.imwrite(label_slice_data,
# # osp.join(out_root, 'label', osp.basename(img).replace('.tif', '_' + str(i) + '.png')))
# print(f"影像{img}已输出，输出路径为{osp.join(out_root, 'label', osp.basename(img).replace('.tif', '_' + str(i) + '.tif'))}")
for i, img in enumerate(img_list):
print(f"""第{i + 1}景影像开始处理，影像名：{os.path.basename(img)}""")
save_8bit_path = img.replace(osp.join(root, prefixs[0]), save_8bit_dir)
if transfor_8bit:
save_8bit_path = save_8bit_path.replace(os.path.splitext(save_8bit_path)[-1], '_8bit.tif')
print(f"""第{ i + 1}景影像开始转8位，影像名：{os.path.basename(img)}""")
_, image_info = imageto8bit(img, save_8bit_path=save_8bit_path)
# data_8bit = data_8bit.astype(np.uint8)
img_data = mmcv.imread(save_8bit_path, backend='tifffile')
else:
img_data = mmcv.imread(img, backend='tifffile')
label_data = mmcv.imread(osp.join(label_root, osp.basename(save_8bit_path).replace('.tif', '_label.tif')),
backend='tifffile')
assert img_data.shape[0] == label_data.shape[0], '影像行数量与标注行数量不匹配'
assert img_data.shape[1] == label_data.shape[1], '影像列数量与标注列数量不匹配'
indexs = slice_index(img_data, slice_size=slice_size)
percent = 0.2
percent_nodata = 0.3
for i, idx in enumerate(indexs):
label_slice_data = slice(label_data, idx, slice_size=slice_size, edge_fill=True, fliter=True,
percent=percent, percent_nodata=percent_nodata)
# print(np.unique(label_slice_data))
if label_slice_data is not None:
mmcv.imwrite(label_slice_data,
osp.join(out_root, 'labels', osp.basename(img).replace('.tif', '_' + str(i) + '.png')))
img_slice_data = slice(img_data, idx, slice_size=slice_size, edge_fill=True, fliter=True,
percent=percent, percent_nodata=percent_nodata)
# print(np.unique(img_slice_data))
# label_slice_data = slice(label_data, idx, slice_size=slice_size, edge_fill=True, fliter=True, percent=percent, percent_nodata=percent_nodata)
if img_slice_data is not None:
mmcv.imwrite(img_slice_data, osp.join(out_root, 'images',
osp.basename(img).replace('.tif', '_' + str(i) + '.tif')))
print(f"影像{img}已输出，输出路径为{osp.join(out_root, 'label', osp.basename(img).replace('.tif', '_' + str(i) + '.tif'))}")