Comments for PaulNovo.us

Comment on Getting Started with Python-InsightToolkit by loli

loli — Thu, 15 Dec 2011 12:49:31 +0000

@paul: That sadly poses a problem for me, as ITK won’t compile with the python bindings enabled on my machine.

I found two workarounds, which both have their problems:
I. Use itk.RelabelComponentImageFilter, as it supports UL input
The problem is here that when more than ushort (~65.000) labels are created, the ones out of range are concatenated
II. Cast to VTK and save as MetaIO image
This solves the problem of I, but one is restricted to the MetaIO format and the execution takes signif. longer

Any hint how I could recompile your pre-compiled package with UL support enabled?

Comment on Getting Started with Python-InsightToolkit by paul

paul — Wed, 30 Nov 2011 21:15:28 +0000

@loli I am not compiling unsigned longs(UL) in these packages. It looks like the WatershedImageFilter has a hard-coded output of UL, so that is why it returns UL. Unfortunately, you would have to recompile the packages with UL turned on, so that all the other filters will also be able to use this data type.

Comment on Getting Started with Python-InsightToolkit by loli

loli — Fri, 25 Nov 2011 15:26:59 +0000

I stumbled the problem that itk.Image.UL2/3 exist and is returned by filters such as itk.WatershedImageFilter. But practically no filter templates exist for this type. Especially frustrating is that neither itk.ImageFileWriter nor itk.RescaleIntensityImageFilter nor the most simple itk.CastImageFilter can handle the UL type.

Any suggestion how I could retrieve my data without having to write my own conversion filter?

Greetings.

Comment on Getting Started with Python-InsightToolkit by paul

paul — Fri, 30 Sep 2011 12:57:39 +0000

@shan I don't have floats wrapped for that type, but it looks like doubles are. If you use ipython you can figure out what types are wrapped and which aren't. See my comment here.

Comment on Getting Started with Python-InsightToolkit by shan

shan — Wed, 28 Sep 2011 01:38:51 +0000

Hi, I am wondering if itk.Mesh has a wrapper. I tried doing as follows
itk.Mesh[itk.F,3]

it say that there is no such template. Or I am not doing it correctly???

Thanks

Comment on Insight Toolkit Ubuntu Repository by Andre

Andre — Fri, 29 Jul 2011 18:20:08 +0000

Great! Using Ipython I still get to know practical tricks. Thanks a lot !

Comment on Insight Toolkit Ubuntu Repository by paul

paul — Fri, 29 Jul 2011 14:31:04 +0000

Are you using ipython? I use ipython’s autocomplete to tell if a filter supports different types.

For example, if you type:
a = itk.GradientImageFilter.
and then hit Tab, you will get a list of its member functions along with this:

itk.GradientImageFilter.IF2FF
itk.GradientImageFilter.IF3FF
itk.GradientImageFilter.ISS2FF
itk.GradientImageFilter.ISS3FF
itk.GradientImageFilter.IUC2FF
itk.GradientImageFilter.IUC3FF
itk.GradientImageFilter.IUS2FF
itk.GradientImageFilter.IUS3FF

It takes a bit to get used to this info, these describe the available templates. IF3FF corresponds to:
itk::GradientImageFilter, float, float>

Comment on Insight Toolkit Ubuntu Repository by Andre

Andre — Fri, 29 Jul 2011 14:17:02 +0000

Can you tell what templating is possible, i.e. does GradientImageFilter support image type [itk.F,3]?

Comment on Getting Started with Python-InsightToolkit by Helen

Helen — Wed, 08 Jun 2011 18:31:47 +0000

Thanks, this was really useful.
Has anyone tried using image registration with python? I’m trying to convert Imageregistration7 example and getting the error:
RuntimeError: /tmp/buildd/insighttoolkit-3.18.0/Code/Numerics/itkRegularStepGradientDescentBaseOptimizer.cxx:197:
itk::ERROR: RegularStepGradientDescentOptimizer(0xad25f0e8): The size of Scales is 6, but the NumberOfParameters for the CostFunction is 4.

Does anyone know how to increase the number of parameters in python?

Comment on Insight Toolkit Ubuntu Repository by paul

paul — Tue, 12 Apr 2011 11:12:25 +0000

@Andrey: When you set the input of one filter with the output of another, you need to use the GetOutput function. So your last line should be subtractImageFilter = itk.SubtractImageFilter[image, image, image].New( Input1 = image, Input2=laplacianImageFilter.GetOutput())