CODEWITHSUNDEEP
simpleitk
Yang Li
Yang Li

Reputation: 1

How to use SimpleITK.WarpImageFilter function in python

I can't find any python documentation on this function.

I tried the following code, but the output is absolutely wrong.

warp = sitk.WarpImageFilter()
warp.SetInterpolator(sitk.sitkLinear)
out = warp.Execute(reference_img, transform_img)

it gives me exactly the reference_img, but not the transformed image.

Have you ever used this function? Could you please help me on this issue?

Upvotes: 0

Views: 1545

Answers (1)

blowekamp
blowekamp

Reputation: 1431

The WarpImageFilter, as with all of the other SimpleITK filters, is rigorously tested in the SimpleITK build infrastructure. The filter wrapping include descriptions of multiple test where the results are validated and the behavior across languages is verified. Additionally in SimpleITK, all filters are automatically tested for consistency on several behaviors which is accomplished though the code generation process SimpleITK has.

The ResampleImageFilter is now the preferred method over the WarpImageFilter. It applies any transform to resample an image, where a DisplacementFieldTransform contains the displacement field image. This is different than the WarpImageFilter which just takes the displacement field.

There seems to be some confusion for the "reference image". Both the resample and the warp filter need to know the output information for the image. This includes the output origin, output size, output spacing and output direction. These can manually be set individually or from a "reference image". There is the ResampleImageFilter.SetRefenceImage() and the WarpImageFilter.SetOutputParametersFromImage methods which do the same thing. the reference image is optional.

Sample usage of the WarpImageFilter utilizing the SimpleITK test data:

img = sitk.ReadImage("cthead1-Float.mha")
dis = sitk.ReadImage("cthead1-dis1.nrrd")
warper = sitk.WarpImageFilter()
warper.SetOutputParameteresFromImage(img)
out = warper.Execute(img,dis)

Here is the same operation done with the ResampleImageFilter. It requires converting the displacement field image to float64 to be used with the transform:

resampler = sitk.ResampleImageFilter()
resampler.SetReferenceImage(img)
dis_tx = sitk.DisplacementFieldTransform(sitk.Cast(dis,sitk.sitkVectorFloat64))
resampler.SetTransform(dis_tx)
out = resampler.Execute(img)

Upvotes: 2

Related Questions