import produces segmentation fault about itkanisotropicdiffusionlbr HOT 7 CLOSED

Hello @thewtex ,
It seems to work now :) thanks a lot!

from itkanisotropicdiffusionlbr.

This

instance = itk.CoherenceEnhancingDiffusionImageFilter[itk.Image[itk.F,2]].New(rescaled_image)

Needs to be

instance = itk.CoherenceEnhancingDiffusionImageFilter[itk.Image[itk.F,2]].New(rescaled_image)
instance.Update()

or the filter is not executed, and the output is not populated. Alternatively, use ITK 5 Pythonic syntax, which determines the required type, runs .Update() and returns the filter's output:

smoothed_image = itk.coherence_enhancing_diffusion_image_filter(rescaled_image)

Also, this does not work:

instance = itk.GetArrayViewFromImage(instance)

because it is generate a view on an image instance (i.e. a look into the image's pixel buffer memory), but then it is overwriting the image instance. The view will not be pointing to anything that exists, which will cause a crash. Instead, use:

instance = itk.GetArrayFromImage(instance)

or the nicer:

instance = itk.array_from_image(instance)

from itkanisotropicdiffusionlbr.

itk.array_from_image(instance)

@thewtex Awesome!! Thank you very much, it really saved me a lot of time.
Thanks again for the fast and helpful answers!

from itkanisotropicdiffusionlbr.

Hello Dimitris @dimitrisl ,

A new version, 1.0.2, built against itk-5.0.0.post1 has been uploaded to PyPI, which should fix your segmentation fault issue.

Please try:

python -m pip install --upgrade itk-anisotropicdiffusionlbr

and let us know how it goes.

from itkanisotropicdiffusionlbr.

hello again @thewtex

it seems that another issue occured.
In [5]: instance = itk.CoherenceEnhancingDiffusionImageFilter(rescaled_image)
Segmentation fault (core dumped)

from itkanisotropicdiffusionlbr.

@dimitrisl could a SSCCE be shared?

from itkanisotropicdiffusionlbr.

@thewtex

import matplotlib.pyplot as plt
import numpy as np
import itk

image = np.random.randn(1517, 1501).astype("float32")
rescaled_image = (image - np.min(image))/np.ptp(image)
rescaled_image = itk.GetImageViewFromArray(rescaled_image)
instance = itk.CoherenceEnhancingDiffusionImageFilter[itk.Image[itk.F,2]].New(rescaled_image)

instance = itk.GetArrayViewFromImage(instance)
plt.imshow(instance) # essentially it breaks on this point here..but this behaviour is not consistent
plt.show()

from itkanisotropicdiffusionlbr.