VT - How to work with transformations

import vt
from timagetk.io.image import _image_from_url

/builds/J-gEBwyb/0/mosaic/timagetk/src/timagetk/components/labelled_image.py:31: TqdmWarning: IProgress not found. Please update jupyter and ipywidgets. See https://ipywidgets.readthedocs.io/en/stable/user_install.html
  from tqdm.autonotebook import tqdm

tmp_path = _image_from_url('https://zenodo.org/record/7149218/files/090223-p58-flo-top.lsm',
                           hash_value='5548917d3d1490200d0d56cbb31c0d35', hash_method='md5')
im = vt.vtImage(tmp_path)

vxs = im.spacing()  # XYZ sorted voxelsize
print(vxs)

[0.2003195434808731, 0.2003195434808731, 1.0]

Create transformations

Identity transformation

id_trsf = vt.create_trsf(params=' -trsf-type rigid -identity')
print(id_trsf)
print(id_trsf.copy_to_array())

vtTransformation : { 
 type    : RIGID_3D, 
 unit    : real, 
 }
[[1. 0. 0. 0.]
 [0. 1. 0. 0.]
 [0. 0. 1. 0.]
 [0. 0. 0. 1.]]

Random linear transformation

rand_linear_trsf = vt.create_trsf(im, params=' -trsf-type rigid -value random -srandom 12 -angle-range 0.0 0.1')
print(rand_linear_trsf)
print(rand_linear_trsf.copy_to_array())

vtTransformation : { 
 type    : RIGID_3D, 
 unit    : real, 
 }
[[ 0.99779585  0.06530346 -0.01178573  3.76788433]
 [-0.06579991  0.99666279 -0.04830798 -7.40149029]
 [ 0.00859172  0.04897701  0.99876295 -3.77669926]
 [ 0.          0.          0.          1.        ]]

Random non-linear transformation

We cannot access the non-linear transformation in VT, so we have to save the transfo with its write method and read it with vtImage to make sure we have indeed a vectorfield transformation of the right shape!

rand_vf_trsf = vt.create_trsf(im, params=" -trsf-type vectorfield -value sinus3D")

from tempfile import NamedTemporaryFile
with NamedTemporaryFile() as f:
    fname = f.name+'.inr'
    rand_vf_trsf.write(fname)

print(fname)

/tmp/tmpfl9aygo7.inr

trsf_im = vt.vtImage(fname)
print(trsf_im.copy_to_array().shape)

(59, 460, 460, 3)

Order of the last axe? XYZ or ZYX ?

Apply transformations

Isotropic voxel

# Use the `-isotropic-voxel` string parameter to performs isometric resampling of the image
id_trsf = vt.create_trsf(params=' -trsf-type rigid -identity')
iso_image = vt.apply_trsf(im, id_trsf, None, params=f'-isotropic-voxel {min(vxs)} -interpolation linear')
print(iso_image)

vtImage : { 
 type    : unsigned char, 
 shape  : [460, 460, 295], 
 spacing : [0.200320, 0.200320, 0.200320], 
 }

Linear transformations

# Apply the random linear transformation with linear interpolation
reg_image = vt.apply_trsf(im, rand_linear_trsf, None, params='-interpolation linear')
print(reg_image)

vtImage : { 
 type    : unsigned char, 
 shape  : [460, 460, 59], 
 spacing : [0.200320, 0.200320, 1.000000], 
 }

# Add isometric image as template to resample and apply the random linear transformation with linear interpolation
iso_reg_image = vt.apply_trsf(im, rand_linear_trsf, iso_image, params='-interpolation linear')
print(iso_reg_image)

vtImage : { 
 type    : unsigned char, 
 shape  : [460, 460, 295], 
 spacing : [0.200320, 0.200320, 0.200320], 
 }

Non-linear transformation

# Apply the random vectorfield transformation with linear interpolation
reg_image = vt.apply_trsf(im, rand_vf_trsf, None, params='-interpolation linear')
print(reg_image)

vtImage : { 
 type    : unsigned char, 
 shape  : [460, 460, 59], 
 spacing : [0.200320, 0.200320, 1.000000], 
 }

# Add isometric image as template to resample and apply the random vectorfield transformation with linear interpolation
iso_reg_image = vt.apply_trsf(im, rand_vf_trsf, iso_image, params='-interpolation linear')
print(iso_reg_image)

None

BAL_ResampleImage: result image and vector field transformation have different dimensions
	 result image   = [460 460 295]
	 transformation = [460 460 59]
API_applyTrsf: unable to compute resampling
API_applyTrsf(vtImageBridge*, vtTransformationBridge*, vtImageBridge*, const std::string&, const std::string&): unable to resample image

Providing a template (or ref) image to vt.apply_trsf does not work!