[Insight-users] Questions about computing distance map using FastMarchingImageFilter

[Dan Mueller]

Hi Siyi,

I would like to help you with the issue you reported, however I am
hindered by the following reasons:

1. I don't know what operation system you are using (Windows, Mac,
Linux, Solaris, etc.)
2. I don't know the version of the components you are using (CMake,
ITK, compiler)
3. I don't know the context of the task you are trying to achieve (are
you computing the distance map of the entire image, or only a small
border around specific objects?)
4. There are no (small) screenshots attached to your email to evaluate
what you mean by "the result does not look like a correct distance
map"
5. The source code snippet you provided does not compile or run on my
machine (no cmake file, not complete minimal example, no input image).

Please help me help you. If you want a fast reply, please make it as
easy as possible for someone on the list to quickly experience the
issue you are facing by providing the above information.

I can take a stab in the dark, and give you a few general pointers:

a. The FastMarchingImageFilter is not the optimal way of computing a
distance map of an entire image. In my experience the
SignedMaurerDistanceImageFilter is the fastest (I think there is even
a multi-threaded version floating around on the Insight Journal). That
said, fast marching *may* be useful (ie. faster) for computing a local
distance map, around the borders of an object (ie. by setting the
stopping value to a small value). I don't know the full context of
your problem, so I'm not sure if these comments are helpful.
b. Performance can generally be improved (sometimes by 10-100 times)
by compiling your application in "Release" mode. Have you compiled
your application using optimizations?

Regards,

Dan

2009/10/2 siyi ding <no16sylvan at hotmail.com>
>
>
> Dear all,
>
> I am trying to use the FastMarchingImageFilter to compute the distance map of a canny edge image. I could get correct distance map when I follow the example - just give one seed point. However, when I add all the none zero points(about 9000 points) in the canny image as seeds, the program becomes very slow and the result does not look like a correct distance map. Is there any one who could help about this?
>
> Thanks a lot.
>
> //////////////////////////////////////////////////////
>  seeds->Initialize();
>  NodeType node;
>  const double seedValue = 0.0;
>  int num_seeds = 0;
> // If we only use one seed as below, we get the correct distance map of that point.
>  /*
>  OutputImageType::IndexType  seedPosition;
>  seedPosition[0] = 100;
>  seedPosition[1] = 100;
>  node.SetValue( seedValue );
>  node.SetIndex( seedPosition );
>  seeds->InsertElement(num_seeds++, node );
>
> */
>
> // If we use all the nonzeros points in the canny edge image as seeds,we can not get correct distance map and the program is slow.
>
>  for (int i = 0; i< (image->w());i++)
>   for (int j = 0; j< (image->h());j++)
>   {
>    OutputImageType::IndexType  seedPosition;
>    seedPosition[0] = i;
>    seedPosition[1] = j;
>
>    if ( (CannyFilter->GetOutput()->GetPixel(seedPosition))>0)
>    {
>     node.SetValue( seedValue );
>     node.SetIndex( seedPosition );
>     seeds->InsertElement(num_seeds++, node );
>    }
>   }
>  fastMarching->SetTrialPoints(  seeds  );
>  fastMarching->SetSpeedConstant(1.0);
>  fastMarching->SetOutputSize(  CannyFilter->GetOutput()->GetBufferedRegion().GetSize() );
>  fastMarching->SetOutputOrigin ( CannyFilter->GetOutput()->GetOrigin());
>  fastMarching->SetOutputSpacing ( CannyFilter->GetOutput()->GetSpacing());
>
> const double stoppingTime = max(image->w(),image->h());
>  fastMarching->SetStoppingValue(  stoppingTime  );
>
> FileWriterType::Pointer writer4 = FileWriterType::New();
>     writer4->SetFileName("distance.jpg");
>     writer4->UseCompressionOn();
>
>  rescaleFilter->SetOutputMinimum(   0 );
>  rescaleFilter->SetOutputMaximum( 255 );
>  rescaleFilter->SetInput(  fastMarching->GetOutput()  );
>  floatToUnsignedCharFilter ->SetInput( rescaleFilter->GetOutput() );
>  writer4->SetInput( floatToUnsignedCharFilter->GetOutput() );
>
>  try
>     {
>   writer4->Update();
>   writer4->DebugOn();
>     }
>     catch (itk::ExceptionObject &e)
>     {
>     std::cerr << e << std::endl;
>     }
>
> Best,
> Siyi Ding
>
>
>
> Best,
> Sylvia Siyi Ding
>
>
>
>
> > Date: Thu, 1 Oct 2009 13:41:21 -0400
> > From: mike.jackson at bluequartz.net
> > To: insight-users at itk.org
> > Subject: [Insight-users] Threaded Gaussian Filter
> >
> > Anyone have a threaded version of the Gaussian Filter? I am currently
> > using the example code for a Gaussian filter from the ITK distribution
> > and for the image size I am throwing at it, the iteration loop takes a
> > long time because it performs the filter serially on every pixel.
> >
> > Thanks
> > _________________________________________________________
> > Mike Jackson mike.jackson at bluequartz.net
> > BlueQuartz Software www.bluequartz.net
> > Principal Software Engineer Dayton, Ohio
> > _____________________________________
> > Powered by www.kitware.com
> >
> > Visit other Kitware open-source projects at
> > http://www.kitware.com/opensource/opensource.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-users
>
> ________________________________
> Lauren found her dream laptop. Find the PC that’s right for you.
> _____________________________________
> Powered by www.kitware.com
>
> Visit other Kitware open-source projects at
> http://www.kitware.com/opensource/opensource.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-users
>