Rotate image to napari camera view

rotate.py

"""
Set napari to 3-D mode and rotate it to the desired view

The output size of the rotated image need to be specified
to it will fit into new image. If the image in not spatially
isotropic it should probably be downsampled (likely in y and x).
I use skimage.transform.downscale_local_mean.

Variables
=========

dstack : array
  The image array displayed in napari to be rotated
 
"""

import numpy as np
import scipy.ndimage as ndi
from scipy.spatial.transform import Rotation as R

camera_initial = np.array((-1, 0, 0))   ## napari initial camera vector (always this)
up_initial = np.array((0, -1, 0)) ## initial up vector

camera_rot = np.array(viewer.camera.view_direction)  ## the camera direction after rotation (into screen)
up_rot = viewer.camera.calculate_nd_up_direction(3, (0, 1, 2)) # the up direction after rotation (screen up)

v_initial = np.array([camera_initial, up_initial])  # list of inital vectors
v_rotated = np.array([camera_rot, up_rot])   # list of rotated vectors

rot, rssd = R.align_vectors(v_initial, v_rotated, return_sensitivity=False)
rot_matrix = rot.as_matrix()

## get the center of the image array
center =(np.array(dstack[4].shape) - 1)/2

"""calculate the offset, this is the point the
This is the point the image will be rotated around.
"""
offset = center - rm.T @ np.array((250, 250, 250))

rxmaxlist = list()

### loop over the channels, assumes channel_axis=0
for c in dstack:
    _rx = ndi.affine_transform(c, rm.T, order=2,
                            output_shape=(500, 500, 500),
                            offset=offset)
    _rxmax = _rx.max(axis=0)
    rxmaxlist.append(_rxmax)
    
rxmax = np.stack(rxmaxlist)  ## the output image

"""
After the rotation is finished, add it to napari, then push the home button.
The home view of the new image should look like the rotated view
of the original image.
"""