Elastix¶
This notebooks show very basic image registration examples with on-the-fly generated binary images.
In [1]:
import itk
import numpy as np
import matplotlib.pyplot as plt
Image generators¶
In [2]:
def image_generator(x1, x2, y1, y2, mask=False, artefact=False):
if mask:
image = np.zeros([100, 100], np.uint8)
else:
image = np.zeros([100, 100], np.float32)
image[y1:y2, x1:x2] = 1
if artefact:
image[-10:, :] = 1
image = itk.image_view_from_array(image)
return image
Masked Registration Test¶
In [4]:
# Create rigid transformed test images with artefact
fixed_image = image_generator(25, 75, 25, 75, artefact=True)
moving_image = image_generator(1, 51, 10, 60, artefact=True)
# Create mask for artefact
fixed_mask = image_generator(0, 100, 0, 90, mask=True)
moving_mask = image_generator(0, 100, 0, 90, mask=True)
# Import Default Parameter Map
parameter_object = itk.ParameterObject.New()
default_rigid_parameter_map = parameter_object.GetDefaultParameterMap('rigid')
parameter_object.AddParameterMap(default_rigid_parameter_map)
# Call registration function
result_image, result_transform_parameters = itk.elastix_registration_method(
fixed_image, moving_image,
parameter_object,
fixed_mask=fixed_mask,
moving_mask=moving_mask)
Masked Registration Test Visualization¶
In [5]:
%matplotlib inline
# Plot images
fig, axs = plt.subplots(1,5, sharey=True, figsize=[30,30])
plt.figsize=[100,100]
axs[0].imshow(fixed_mask)
axs[0].set_title('Fixed Mask', fontsize=30)
axs[1].imshow(moving_mask)
axs[1].set_title('Moving Mask', fontsize=30)
axs[2].imshow(fixed_image)
axs[2].set_title('Fixed', fontsize=30)
axs[3].imshow(moving_image)
axs[3].set_title('Moving', fontsize=30)
axs[4].imshow(result_image)
axs[4].set_title('Result', fontsize=30)
plt.show()
