[Insight-users] What is a good image format for ITK?

[hhiraki@lab.nig.ac.jp]

Hi Mathieu,

I tried a cvs version. It's on a pentium3 linux box with gcc-3.2.
Thanks to Stephen R. Aylward, MetaImageIO had already fixed. The 
other two can be reproduced with Examples/IO/ImageCastReadWrite.cxx
and Examples/Data/BrainT1Slice.png.


On VTKImageIO, as the default format had been changed to binary, 
ImageCastReadWrite.cxx had to be changed to reproduce the flaw.
Fow example:

--- ImageReadCastWrite.cxx~	Wed May 28 14:49:52 2003
+++ ImageReadCastWrite.cxx	Wed May 28 15:06:27 2003
@@ -205,6 +205,8 @@
   try 
     { 
     writer->Update(); 
+    writer->GetImageIO()->SetFileTypeToASCII();
+    writer->Update(); 
     } 
   catch( itk::ExceptionObject & err ) 
     { 

Then try:

$ ImageReadCastWrite BrainT1Slice.png foobar.vtk

The result file foobar.vtk causes an error in VTK:

$ vtk
% vtkStructuredPointsReader reader
reader
% reader SetFileName foobar.vtk
% reader Update
Generic Warning: In /tmp/src/VTK/IO/vtkDataReader.cxx, line 833
Error reading ascii data!


On AnalyzeImageIO, convert the image through AnalyzeIO:

$ ImageReadCastWrite BrainT1Slice.png foobar.hdr
$ ImageReadCastWrite foobar.hdr foobar.png

Then, the grayscale of the result foobar.png was overflowed at 
the value 128. 


By the way, I think signed 8bit integer pixel type must be
explicit "signed char" rather than platform-dependent "char".

Regards,

Hideaki Hiraki


Mathieu Malaterre wrote:
> Hideaki Hiraki,
> 
> There has been quite a lot of bug fix since ITK1.2 in all of the ImageIO 
> you are talking about:
> 
> http://public . kitware . com/cgi-bin/cvsweb . cgi/Insight/Code/IO/itkVTKImageIO . cxx?cvsroot=Insight
> http://public . kitware . com/cgi-bin/cvsweb . cgi/Insight/Code/IO/itkAnalyzeImageIO . cxx?cvsroot=Insight
> http://public . kitware . com/cgi-bin/cvsweb . cgi/Insight/Code/IO/itkMetaImageIO . cxx?cvsroot=Insight
> 
> Could you please try a CVS version of ITK ? And if this is still not 
> working could you give us a short example on how to reproduce each bugs, 
> thanks ?
>