|
1
|
|
|
from torchio.data.image import ScalarImage |
|
2
|
|
|
from typing import Optional |
|
3
|
|
|
import torch |
|
4
|
|
|
from ....data.subject import Subject |
|
5
|
|
|
from ...intensity_transform import IntensityTransform |
|
6
|
|
|
|
|
7
|
|
|
|
|
8
|
|
|
class Clamp(IntensityTransform): |
|
9
|
|
|
"""Clamp intensity values into a range :math:`[a, b]`. |
|
10
|
|
|
|
|
11
|
|
|
For more information, see :func:`torch.clamp`. |
|
12
|
|
|
|
|
13
|
|
|
Args: |
|
14
|
|
|
out_min: Minimum value :math:`a` of the output image. If ``None``, the |
|
15
|
|
|
minimum of the image is used. |
|
16
|
|
|
out_max: Maximum value :math:`b` of the output image. If ``None``, the |
|
17
|
|
|
maximum of the image is used. |
|
18
|
|
|
|
|
19
|
|
|
Example: |
|
20
|
|
|
>>> import torchio as tio |
|
21
|
|
|
>>> ct = tio.ScalarImage('ct_scan.nii.gz') |
|
22
|
|
|
>>> HOUNSFIELD_AIR, HOUNSFIELD_BONE = -1000, 1000 |
|
23
|
|
|
>>> clamp = tio.Clamp(out_min=HOUNSFIELD_AIR, out_max=HOUNSFIELD_BONE) |
|
24
|
|
|
>>> ct_clamped = clamp(ct) |
|
25
|
|
|
""" |
|
26
|
|
|
def __init__( |
|
27
|
|
|
self, |
|
28
|
|
|
out_min: Optional[float] = None, |
|
29
|
|
|
out_max: Optional[float] = None, |
|
30
|
|
|
**kwargs |
|
31
|
|
|
): |
|
32
|
|
|
super().__init__(**kwargs) |
|
33
|
|
|
self.out_min, self.out_max = out_min, out_max |
|
34
|
|
|
self.args_names = 'out_min', 'out_max' |
|
35
|
|
|
|
|
36
|
|
|
def apply_transform(self, subject: Subject) -> Subject: |
|
37
|
|
|
for image in self.get_images(subject): |
|
38
|
|
|
self.apply_clamp(image) |
|
39
|
|
|
return subject |
|
40
|
|
|
|
|
41
|
|
|
def apply_clamp(self, image: ScalarImage) -> None: |
|
42
|
|
|
image.set_data(self.clamp(image.data)) |
|
43
|
|
|
|
|
44
|
|
|
def clamp(self, tensor: torch.Tensor) -> torch.Tensor: |
|
45
|
|
|
return tensor.clamp(self.out_min, self.out_max) |
|
46
|
|
|
|
fepegar /
torchio
