[vtk-developers] TestSobel2D-image dashboard failures, anyone know anything about this?

[Mathieu Coursolle]

Hi,

You were right, TestSobel2D started to fail after the 2006/05/03 change
of the DICOMAppHelper.cxx file. The change was good, but it made the spacing
values to be set individually. They were set to the same value before. It then
made visible a problem that was already there.

In fact, the string we want to convert is the following:

007.812500E-01\07.812500E-01

On little endians, it is read as follow:

this->PixelSpacing[0] = 0.78125
this->PixelSpacing[1] = 0.78125

However, on Big endian, the string is read as follow:

.7005218-E000\1018.7005210-E

which produces:

this->PixelSpacing[0] = 0.7005218
this->PixelSpacing[1] = 1018.7005210

You'll notice that the Big endian version of the string
is swapped by blocks of 4 bytes.

In the ReadNextRecord method of the DICOMParser class, the Pixel spacing tag
is set to be swapped.

That tag being a string, does it really need to be swapped? Is the
spacing tag always an
ASCII string in the DICOM files?

If I comment the VR_FL case from the switch case statement which
determine what need
to be swapped, TestSobel2D then pass on both intel and PPC.

I attached a patch, for which all tests passed on both intel and PPC
with TCL enabled.
(Except for TestImageTracerWidget on my machine, but this is some other
issue).

Thanks.

Mathieu 


>On 7/9/07, Sebastien BARRE <sebastien.barre at kitware.com> wrote:
>>
>>  At 7/6/2007 07:59 PM, Bill Lorensen wrote:
>>
>> I see a change to DICOMAppHelper.cxx
>> http://public.kitware.com/cgi-bin/viewcvs.cgi/Utilities/DICOMParser/
>DICOMAppHelper.cxx.diff?cvsroot=VTK&r1=1.25&r2=1.26
>>
>
>Sean, can you confirm that applying the reversed (patch -R) patch you
>get a clean dashboard ?
>
>>  Now to be honest I think the DICOM code in VTK should be yanked and
>> replaced with support for GDCM. The DICOM protocol is too complex and too
>> time consuming to be maintained in different libraries any longer, and GDCM
>> has been used in ITK (or internally at Kitware) with good success so
far, if
>> I'm correct.
>
>GDCM is a pretty good choice indeed ;-P
>As a side note, it took me quite a while to get GDCM to compile on the
>mulitple compilers supported by ITK. I think it will be an even
>greater challenge to get it supported by all compilers VTK supports...
>Not mentionning the jpeg8/12/16/j2k library addition
>
>-- 
>Mathieu
>_______________________________________________
>vtk-developers mailing list
>vtk-developers at vtk.org
>http://www.vtk.org/mailman/listinfo/vtk-developers
>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: TestSobel2D_PatchFile.txt
URL: <http://public.kitware.com/pipermail/vtk-developers/attachments/20070710/790ec6dd/attachment.txt>