[Insight-developers] Re: Adaptor update mechanism

[Karthik Krishnan]

Hi Jim,

Thank you very much for looking at it.

The changes I was thinking of in the last mail were incorrect. I think 
I've found it now though.

First, I re-ran the example on Linux and on VS71 and ImageAdaptor2 
crashes if reader->Update() is not called prior to adaptor->SetImage(). 

As you mentioned, the m_BufferedRegion was not being set correctly, 
because the ImageAdaptor calls SetBufferedRegion() only in the function 
SetImage(). So, if a pipeline is set-up prior to execution, I don't see 
how the buffered region could be propagated correctly.

I did fix it though by adding SetBufferedRegion( 
m_Image->GetBufferedRegion() ) right after the m_Image updates its 
output data.

So  ImageAdaptor::UpdateOutputData() would be :

ImageAdaptor<TImage , TAccessor>::UpdateOutputData()
{
  // call the superclass' method first, then delegate
  Superclass::UpdateOutputData();

  // delegation to internal image
  m_Image->UpdateOutputData();
  SetBufferedRegion( m_Image->GetBufferedRegion() );
}

That fixes it, but I am not sure if it is the correct thing to do.

Thanks
regards
karthik

PS: Here were the command line params I used to run ImageAdaptor2 (with 
nightly CVS source)
~/work/ITK/binaries/Insight/Nightly/bin/ImageAdaptor2 
~/work/ITK/src/Insight/Nightly/Examples/Data/VisibleWomanEyeSlice.png 
redEye.png


Miller, James V (Research) wrote:

>Karthik, 
>
>I'd have to look at this in more detail.  Your suggested changes do not seem
>correct to me.  
>
>ImageAdaptor2 works fine for me if I remove the call to reader->Update().
>(Using .Net 2003).
>
>The LargestPossibleRegion and the RequestedRegion should have been propagated
>through the adaptered already.  If there is a problem, it might be somewhere 
>where m_BufferedRegion is accessed directly instead of via an access method.
>
>The only place I see this happening is in ComputeOffsetTable(). We could have
>that method call GetBufferedRegion().  However, I cannot duplicate the error 
>you are getting.
>
>
>Jim
>
>
>-----Original Message-----
>From: Karthik Krishnan [mailto:Karthik.Krishnan at kitware.com]
>Sent: Tuesday, May 31, 2005 5:22 PM
>To: Insight-developers (E-mail); Luis Ibanez; Miller, James V (Research)
>Subject: Adaptor update mechanism
>
>
>Hi,
>
>The adaptors don't seem to propagate information when plugged into the 
>pipeline. For example ImageAdaptor2 example segfaults if 
>reader->Update() is commented out.
>
>I believe that  the Update() call in ImageAdaptor.txx should be
>
>//----------------------------------------------------------------------------
>template <class TImage, class TAccessor >
>void
>ImageAdaptor<TImage , TAccessor>
>::Update()
>{
>  Superclass::Update();
> 
>  Superclass::SetLargestPossibleRegion( 
>m_Image->GetLargestPossibleRegion() );
>  Superclass::SetBufferedRegion( m_Image->GetBufferedRegion() );
>  Superclass::SetRequestedRegion( m_Image->GetRequestedRegion() );
>  m_Image->Update();
>}
>
>instead of what it is now. Please correct me if I am mistaken. I don't 
>want to commit incorrect code since I am not familiar enough with the 
>pipeline mechanism.
>
>Thanks
>Regards
>Karthik
>
>
>  
>