[Insight-users] SignedDanielssonDistanceMapImageFilter

Luis Ibanez luis.ibanez at kitware.com
Wed Dec 10 15:49:39 EST 2008


Hi chensiqi


   Here is the philosophical question that you must
   answer before your proceed further.


      You want to get the distance to "what" ?


   Do you want the distance to the points ?
   Do you want the distance to the boundary ?


   From a set of points located in a boundary is
   *very difficult* to find the inside of an object.

   Do you have a topology connecting these points ?
   e.g. a triangulated surface ?

   How did you get these points ?
   It may be easier to work with the data preceeding
   your point extraction.



      Regards,



          Luis



==================
chensiqi wrote:
> 
> Hi, Luis.
>  
> Thanks for the clarification.
> My purpose is to build a signed distance function given a point clouds 
> of a shape boundary. I know there are several options to do this. If the 
> only input are those points, which is the best, fast and easiest way to 
> do this? I know there are several options in ITK.
> 1.  FastMarching. 
> 2.  Reinitiaze
> 3.  DistanceMap (Danielsson).
>  
> Another question would be regarding smart pointer and raw pointer.  Is 
> it OK to return a smart pointer from a function?
> E.g. I have this code:
> //////////////////////////////////////////////////////////////////////
> main.cpp
> typedef itk::Image< unsigned char, 2 > BinaryImageType;
> typedef itk::Image< float , 2> LSVImageType;
> LSVImageFilter::Pointer LSVImage =  SDF(contour) // where the contour is 
> binary image  
> SDF.cpp
> LSVImageType::Pointer SDF(BinaryImageType::Pointer contour)
> {
>        typedef itk::SignedDanielssonDistanceMapImageFilter < 
> BinaryImageFilter, LSVImageFilter > FilterType;
>        FilterType::Pointer filter = FilterType::New(); 
>        filter ->SetInput ( contour ); // contour is the binary image 
>        return filter->GetOutput();
> } 
> 
> 
> //////////////////////////////////////////////////////////////////////
>  
>  
> Thanks for any input
> Siqi
>  
> 
>  > Date: Wed, 10 Dec 2008 15:16:08 -0500
>  > From: luis.ibanez at kitware.com
>  > To: pidanchen at hotmail.com
>  > CC: insight-users at itk.org
>  > Subject: Re: [Insight-users] SignedDanielssonDistanceMapImageFilter
>  >
>  >
>  > Hi Chensiqi,
>  >
>  > You are misinterpreting the behavior of this filter.
>  >
>  > If you have a binary image, where
>  >
>  > * The foreground object is set to 1 and
>  > * The background object is set to 0
>  >
>  >
>  > Then the SignedDanielssonDistanceMapImageFilter will compute
>  > the distances to the *transition* region between the 0s and
>  > the 1s.
>  >
>  > All pixesl in the 0s region will have positive distances to
>  > that boundary, while all pixels in the 1s region will have
>  > negative distances to that boundary.
>  >
>  > In your case you are feeding an input image where the *boundary*
>  > points are set to 1. Therefore, there is no "inside" to the
>  > object.
>  >
>  > E.g. if you are feeding into this distance filter an image with
>  > a thin ring, you are not going to find the "inside" of the ring.
>  >
>  > You should feed this filter with an image containing a filled
>  > circle of radius R. Where all the points at locations r < R will
>  > be set to 1s. Only in that case will the distance map be able to
>  > find the "inside" of the object.
>  >
>  >
>  > Regards,
>  >
>  >
>  > Luis
>  >
>  >
>  > ---------------
>  > chensiqi wrote:
>  > > Hi, Ghassan
>  > >
>  > > Thanks for the advice. I flip the input.
>  > > The result now is both inside and outside are positive :(
>  > >
>  > > I use float type as the filteroutput instead of "unsigned short" in 
> the
>  > > example. I also tried unsigned short, the result is the same.
>  > >
>  > > Here is part of my code.
>  > > 
> ////////////////////////////////////////////////////////////////////////////////////////////////////
>  > > typedef itk::Image< unsigned char, 2 > BinaryImageType;
>  > > typedef itk::Image< float , 2> LSVImageType;
>  > > typedef itk::SignedDanielssonDistanceMapImageFilter < 
> BinaryImageFilter,
>  > > LSVImageFilter > FilterType;
>  > > FilterType::Pointer filter = FilterType::New();
>  > >
>  > > filter ->SetInput ( contour ); // contour is the binary image with
>  > > BinaryImageType, 1 at boundary points, and 0 elsewhere.
>  > >
>  > > LSVImageFilter::Pointer LSVImage = filter->GetDistanceMap();
>  > >
>  > > Display (LSVImage) // this is a function that displays the image.
>  > > 
> //////////////////////////////////////////////////////////////////////////////////////////////////////
>  > >
>  > > I notice when I cout<<LSV_image->GetBufferedRegion().GetSize() , the
>  > > output is [0,0]. I guess there is something wrong in my code that
>  > > causes this weird output size. I can display LSVImage, although both
>  > > inside and outside region are postive.
>  > >
>  > >
>  > > Thanks for any input
>  > > Siqi
>  > >
>  > > > CC: insight-users at itk.org
>  > > > From: hamarneh at gmail.com
>  > > > To: pidanchen at hotmail.com
>  > > > Subject: Re: [Insight-users] SignedDanielssonDistanceMapImageFilter
>  > > > Date: Tue, 9 Dec 2008 20:45:19 -0800
>  > > >
>  > > > Hello Siqi,
>  > > >
>  > > > You write: "The input is only a binary image with boundary points at
>  > > > 0, and all other points at 1."
>  > > > However, according to he documentation: "As a convention, the 
> distance
>  > > > is evaluated from the boundary of the ON pixels."
>  > > > (from
>  > > 
> http://www.itk.org/Doxygen/html/classitk_1_1SignedDanielssonDistanceMapImageFilter.html 
> 
>  > >
>  > > > )
>  > > > So, your input image should have boundary pixels set to 1 and all
>  > > > other points to 0 (and not the other way around).
>  > > > HTH,
>  > > > /Ghassan
>  > > >
>  > > >
>  > > >
>  > > > On 9-Dec-08, at 7:24 PM, chensiqi wrote:
>  > > >
>  > > > >
>  > > > > Hi, ITKers
>  > > > >
>  > > > > I got two ques tions regarding 
> SignedDanielssonDistanceMapImageFilter.
>  > > > >
>  > > > > 1. The result I have is not "signed", both the inside and outside
>  > > > > distance values are negative. The input is only a binary image with
>  > > > > boundary points at 0, and all other points at 1. I was wondering
>  > > > > how I was wondering how SignedDanielssonDistanceMapImageFilter
>  > > > > determine the region of inside and outside.
>  > > > >
>  > > > > 2. In Examples of SignedDanielssonDistanceMapImageFilter. cxx, I
>  > > > > notice that the OutputImageType of filter is unsigned short, 
> which I
>  > > > > don't quite understand. The output should contains both negative
>  > > > > and positive distance, why use "unsigned" here?
>  > > > >
>  > > > > Thanks
>  > > > > Siqi
>  > > > >
>  > > > > MSN保护盾,十八般武艺保障MSN安全! 现
>  > > > > 在就下载! _______________________________________________
>  > > > > Insight-users mailing list
>  > > > > Insight-users at itk.org
>  > > > > http://www.itk.org/mailman/listinfo/insight-users
>  > > >
>  > >
>  > >
>  > > 
> ------------------------------------------------------------------------
>  > > MSN热搜榜全新升级,更多排行榜等着你! 立即查看! < 
> http://top.msn.com.cn>
>  > >
>  > >
>  > > 
> ------------------------------------------------------------------------
>  > >
>  > > _______________________________________________
>  > > Insight-users mailing list
>  > > Insight-users at itk.org
>  > > http://www.itk.org/mailman/listinfo/insight-users
> 
> 
> ------------------------------------------------------------------------
> 使用新一代 Windows Live Messenger 轻松交流和共享! 立刻下载! 
> <http://im.live.cn/>


More information about the Insight-users mailing list