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
|
import numpy as np
import PIL.Image
from . import utils
try:
import cv2
except ImportError:
cv2 = None # type: ignore
def _resize_pillow(src, height, width, interpolation):
if interpolation == "linear":
interpolation = PIL.Image.BILINEAR
elif interpolation == "nearest":
interpolation = PIL.Image.NEAREST
else:
raise ValueError("unsupported interpolation: {}".format(interpolation))
if np.issubdtype(src.dtype, np.integer):
dst = utils.numpy_to_pillow(src)
dst = dst.resize((width, height), resample=interpolation)
dst = utils.pillow_to_numpy(dst)
else:
assert np.issubdtype(src.dtype, np.floating)
ndim = src.ndim
if ndim == 2:
src = src[:, :, None]
C = src.shape[2]
dst = np.zeros((height, width, C), dtype=src.dtype)
for c in range(C):
src_c = src[:, :, c]
src_c = utils.numpy_to_pillow(src_c)
dst[:, :, c] = src_c.resize((width, height), resample=interpolation)
if ndim == 2:
dst = dst[:, :, 0]
return dst
def _resize_opencv(src, height, width, interpolation):
if interpolation == "linear":
interpolation = cv2.INTER_LINEAR
elif interpolation == "nearest":
interpolation = cv2.INTER_NEAREST
else:
raise ValueError("unsupported interpolation: {}".format(interpolation))
dst = cv2.resize(src, (width, height), interpolation=interpolation)
return dst
def resize(
src,
height=None,
width=None,
interpolation="linear",
backend="auto",
):
"""Resize image.
Parameters
----------
src: numpy.ndarray, (H, W) or (H, W, C)
Input image.
height: int, optional
Height of image. If not given,
the image is resized based on width keeping image ratio.
width: int, optional
Width of image. If not given,
the image is resized based on height keeping image ratio.
interpolation: str
Resizing interpolation (default: 'linear').
'linear':
Linear interpolation.
'nearest':
Interpolate with the nearest value.
backend: str
Resizing backend (default: 'auto').
'pillow':
Pillow is used.
'opencv':
OpenCV is used.
Returns
-------
dst: numpy.ndarray
Resized image.
"""
if not isinstance(src, np.ndarray):
raise TypeError("src type must be numpy.ndarray")
if backend == "auto":
backend = "pillow" if cv2 is None else "opencv"
src_height, src_width = src.shape[:2]
if isinstance(width, float):
scale_width = width
width = int(round(scale_width * src_width))
if isinstance(height, float):
scale_height = height
height = int(round(scale_height * src_height))
if height is None:
assert width is not None
scale_height = 1.0 * width / src_width
height = int(round(scale_height * src_height))
if width is None:
assert height is not None
scale_width = 1.0 * height / src_height
width = int(round(scale_width * src_width))
if backend == "pillow":
dst = _resize_pillow(src, height, width, interpolation)
elif backend == "opencv":
dst = _resize_opencv(src, height, width, interpolation)
else:
raise ValueError("unsupported backend: {}".format(backend))
return dst
|
