point set conversion #13
Conversation
thewtex
commented
Oct 16, 2025
thewtex
commented
- ENH: Bump version to 0.4.0
- ENH: Initial points_layer_from_point_set, point_set_from_points_layer
- DOC: Describe point set conversion, expand on image conversion
- DOC: Add support for applying points layer rotate
|
Closes #2 |
thewtex
force-pushed
the
point-set-conversion
branch
from
4ed930d to
8654b2b
Compare
| # The layer will have: | ||
| # - image_layer.scale = image spacing (in reverse order for NumPy) | ||
| # - image_layer.translate = image origin (in reverse order for NumPy) | ||
| # - image_layer.rotate = image direction matrix (in reverse order for NumPy) |
There was a problem hiding this comment.
image_layer.rotate is the transpose of ITK's image direction matrix, right? (Rather than "in reverse order"). At least, that's what I understand from your 7e48ef0
There was a problem hiding this comment.
There is the issue of the order elements of the columns / rows (which I am trying to indicate with "in reverse order", and which I assume is correct), and there is the issue of how the matrix is applied, including whether a transpose is needed. The latter is a separate issue, and I think it is harder to make an assumption, so I would not like to make assertions here ;-).
README.md
Outdated
| layer = viewer.add_image( | ||
| data, | ||
| scale=[2.0, 1.5, 1.5], # anisotropic spacing | ||
| translate=[10.0, 20.0, 30.0], | ||
| metadata={'description': 'My volume'} | ||
| ) | ||
|
|
||
| # Convert to ITK | ||
| image = image_from_image_layer(layer) | ||
|
|
||
| # The ITK image will have: | ||
| # - spacing: coordinates in ITK order, so reversed from napari `scale`: [1.5, 1.5, 2.0] | ||
| # - origin: coordinates in ITK order, so reversed from napari `translate`: [30.0, 20.0, 10.0] | ||
| # - direction: from napari `rotate`. In this case unspecified, so just the identity matrix. |
There was a problem hiding this comment.
Can you possibly extend the example here with a non-identity rotation matrix:
viewer.add_image(
data,
scale=[2.0, 1.5, 1.5], # anisotropic spacing
translate=[10.0, 20.0, 30.0],
rotate=[ ... ],And then possibly mention in the comment (line 100 here) what the ITK direction then would be?
Documentation and test covering napari Image Layer rotate and the itk.Image direction as a NumPy array an itk.Matrix.
| # napari rotate is transposed compared to ITK direction | ||
| assert np.allclose(rotate, np.transpose(np.array(image["direction"]))) |
There was a problem hiding this comment.
Cool, thanks Matt!!! So the napari rotate is just the transpose of the ITK "direction", right?
There was a problem hiding this comment.
Locally I got:
Non-orthogonal slicing is being requested, but is not fully supported. Data is displayed without applying an out-of-slice rotation or shear component.
Traceback (most recent call last):
File "D:\X\Src\M\MyPythonApp\MyPythonApp.py", line 41, in <module>
test_image_from_image_layer_rotate_3d()
File "D:\X\Src\M\MyPythonApp\MyPythonApp.py", line 27, in test_image_from_image_layer_rotate_3d
assert np.allclose(rotate, np.transpose(np.array(image["direction"])))
AssertionError
Could be my fault, still having another look 🤷
There was a problem hiding this comment.
By the way, the assert np.allclose(rotate, np.transpose(np.array(image["direction"]))) looks somewhat redundant, because the next lines of code check image["direction"] as well. Especially:
expected_direction = np.transpose(rotate)
assert np.allclose(direction_array, expected_direction)
There was a problem hiding this comment.
Sorry for the noise, I was running your new unit test on the old itk_napari_conversion version!!! That causes test assert failures, of course 😄
2 participants
