[Insight-developers] Wrong behaviour of ImageFileReader when reading an itk::VectorImage<itk::FixedArray<double, 6>, 3> ?

[Karthik Krishnan]

Benoit:

Do you have a variable number of tensors at each pixel ? If not, You should
simply declare your image as

Image< Vector< DiffusionTensor< double >, N >, 3 >

The whole intent of the VectorImage was to avoid memory fragmentation by
internally storing images of the type
  Image< Array< double >, 3 >
as a contiguous set of N doubles for the array, instead of a pointer to the
data at each pixel and the length of the data at each pixel (itk::Array<
double > has a pointer to the data and an integer to represent the length of
the pixel.). The resulting image will behave just as itk::Image< Array<
double >, 3 > would behave. It will provide faster memory access, consume
less memory and provide a less fragmented memory representation.

By contrast, the compile time classes, ie Vector and DiffusionTensor are
equivalent to a C array, and 0 overhead.

-
karthik


On Mon, Jan 17, 2011 at 9:16 PM, Benoit Scherrer
<benoitscherrer at gmail.com>wrote:

> Ok thanks!
>
> But so there is no way to use:
> typedef itk::VectorImage<itk::DiffusionTensor<double>, 3>
> TensorVectorImage;
> to read a set of tensors at each voxel ?
> ( itk::DiffusionTensor inherits itk::FixedArray)
>
> It would be useful to do for example:
>
> TensorVectorImage tensors;
> tensors[i].GetFractionalAnisotropy()
>
> Thanks
> Benoit
>
>
> On Thu, Jan 13, 2011 at 11:26 PM, Karthik Krishnan <
> karthik.krishnan at kitware.com> wrote:
>
>>
>>
>> On Fri, Jan 14, 2011 at 6:26 AM, Benoit Scherrer <
>> benoitscherrer at gmail.com> wrote:
>>
>>> Hi,
>>>
>>> I wanted to use the type:
>>> typedef itk::VectorImage<itk::FixedArray<double, 6>, 3>
>>> TensorVectorImage;
>>> to make a vector image of tensors.
>>>
>>> Everything compiles well. But when i use the
>>> itk::ImageFileReader<TensorVectorImage>
>>> it seems that things goes wrong. When trying to iterate on the image :
>>>
>>> itk::ImageRegionConstIterator<TensorVectorImage>
>>>  it(image, image->GetLargestPossibleRegion());
>>>  while(!it.IsAtEnd())
>>> {
>>> itk::VariableLengthVector<itk::FixedArray<double, 6> > tensors =
>>> it.Get();
>>>                (...)
>>> ++it;
>>> }
>>>
>>> it compiles but during the execution goes out of the data bounds.
>>>
>>> By looking into ImageFileReader<TOutputImage,
>>> ConvertPixelTraits>::GenerateOutputInformation
>>> I found the lines:
>>>
>>>   if( strcmp( output->GetNameOfClass(), "VectorImage" ) == 0 )
>>>     {
>>>     typedef typename TOutputImage::AccessorFunctorType
>>> AccessorFunctorType;
>>>     AccessorFunctorType::SetVectorLength( output,
>>> m_ImageIO->GetNumberOfComponents() );
>>>     }
>>>
>>> So here m_ImageIO->GetNumberOfComponents() will returns 6 for a simple
>>> tensor file (Dimension
>>> of the VectorImage = 1), instead of 1 (after the offsets computed
>>> in ImageRegionConstIterator are wrong).
>>>
>>> So should ImageFileReader divides m_ImageIO->GetNumberOfComponents() by
>>> the number of components
>>> of TOutputImage::PixelType ? (i manually set it to 1 in the debugger,
>>> it's working. However i didn't tried
>>> with an image with 12 components, ie two tensors)
>>>
>>> Or is it a bad idea to use  itk::VectorImage<itk::FixedArray<double, 6>,
>>> 3>  ?
>>>
>>
>> To make a VectorImage of tensors, you declare it simply as
>> itk::VectorImage< double, 3 > and set the VectorLength on the image to 6.
>>
>> Or you could use itk::Image< FixedArray< double, 6 >, 3 >.
>>
>> The two are equivalent.
>>
>>
>>
>>> Thanks a lot
>>> Benoit
>>>
>>>
>>> _______________________________________________
>>> Powered by www.kitware.com
>>>
>>> Visit other Kitware open-source projects at
>>> http://www.kitware.com/opensource/opensource.html
>>>
>>> Kitware offers ITK Training Courses, for more information visit:
>>> http://kitware.com/products/protraining.html
>>>
>>> Please keep messages on-topic and check the ITK FAQ at:
>>> http://www.itk.org/Wiki/ITK_FAQ
>>>
>>> Follow this link to subscribe/unsubscribe:
>>> http://www.itk.org/mailman/listinfo/insight-developers
>>>
>>>
>>
>
>
> --
> / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / /
> / /
> Benoit Scherrer
> Postdoctoral Research Fellow
> Computational Radiology Laboratory,
> Harvard Medical School (Boston)
>
> http://www.crl.med.harvard.edu/
> http://www.BenoitScherrer.com
> / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / / /
> / /
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.itk.org/mailman/private/insight-developers/attachments/20110117/6ad64d87/attachment.htm>