[Insight-users] Possible bug in GaussianSpatialObject

[Ali -]

Julien,

This seems to be one of those hard-to-detect bugs, now we know that it 
is very unlikely that wrapping causes the problem. This is a minimal 
pipeline which can generate the bug:

[itk::GaussianSpatialObject] -> [itk::SpatialObjectToImageFilter] -> 
[itkImageToVTKImageFilter] -> [vtkWriter]

The reason that you have never encountered this problem in ITK is that 
the pipelines normally end with an ITK writer. However, after passing 
the image buffer to a VTK writer, the output file is broken.

One may argue that this is an ItkVtkGlue bug, but the fact that we have 
not observed anything similar in all other image filters suggests that 
there is something different in SpatialObjectToImageFilter. Is it 
possible that some, possibly optional, step is not performed in 
SpatialObjectToImageFilter? If you can try this with ItkVtkGlue, it'd be 
a big help to us.


-Ali


Julien Jomier wrote:
> Hi Ali,
>
> Couple of things:
>
> 1) Change the gaus.GetIndexToObjectTransform().SetTranslation(offset) to
> gaus.GetObjectToParentTransform().SetTranslation(offset)
>
> 2) Add gaus.ComputeObjectToWorldTransform(); after setting the 
> translation. This forces the recomputation of the world transform for 
> the object.
>
> 3) Don't forget to set the size of your image for the 
> SpatialObjectToImageFilter using SetSize().
>
> Here's the code I've been using (C++ version)
>
>   typedef itk::GaussianSpatialObject GSType;
>   GSType::Pointer gsp = GSType::New();
>
>   gsp->SetMaximum(255);
>   gsp->SetRadius(7);
>   typedef itk::Vector VectorType;
>   VectorType v;
>   v[0] = 10;
>   v[1] = 10;
>   gsp->GetObjectToParentTransform()->SetTranslation(v);
>   gsp->ComputeObjectToWorldTransform();
>
>   typedef itk::Image ImageType;
>   typedef itk::SpatialObjectToImageFilter FilterType;
>   FilterType::Pointer filter = FilterType::New();
>   filter->SetInput(gsp);
>   FilterType::SizeType size;
>   size[0]=20;
>   size[1]=20;
>   filter->SetSize(size);
>   filter->Update();
>
>   typedef itk::ImageFileWriter WriterType;
>   WriterType::Pointer writer= WriterType::New();
>   writer->SetFileName("test.jpg");
>   writer->SetInput(filter->GetOutput());
>   writer->Update();
>
> Let me know if that works or not,
>
> Julien
>
> Ali - wrote:
>> Hi Julien,
>>
>> It seems there is this bug in GaussianSpatialObject which causes to mess
>> up the whole spatial object image when the translation transform is
>> applied. Here is an example:
>>
>> itkGaussianSpatialObject2 gaus = new itkGaussianSpatialObject2();
>> gaus.SetMaximum(255);
>> gaus.SetRadius(7);
>> itkVectorD2 offset = new itkVectorD2();
>> offset.SetElement(0, 0.0);
>> offset.SetElement(1, 0.0);
>> gaus.GetIndexToObjectTransform().SetTranslation(offset);
>>
>> (The above code is in java, but can be used as a pseudo code). Notice
>> that whether the offset is zero or not, setting the translation value
>> messes up the gaussian object. Here are the results:
>>
>> Without calling translation: http://i26.tinypic.com/2j4u7pd.jpg
>> With calling translation: http://i31.tinypic.com/iwrd5y.jpg
>>
>> All I wanted to do was to have a gaussian object at the centre of the
>> image. Is there a better way of doing this?
>>
>>
>> -Ali
>>
>>
>>
>>
>> ------------------------------------------------------------------------
>> Miss your Messenger buddies when on-the-go? Get Messenger on your 
>> Mobile! 
>
>



_________________________________________________________________

http://clk.atdmt.com/UKM/go/msnnkmgl0010000007ukm/direct/01/
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.itk.org/pipermail/insight-users/attachments/20080618/37401152/attachment-0001.htm>