[Insight-users] about Level set

[Zachary Pincus]

You're right: I think this is likely due to the narrow band. You 
probably won't see distances too much outside of +/-(narrow band width) 
in the output "distance map", because the whole idea of the narrow band 
is that distances outside of the band aren't calculated.

This really shouldn't be too much of a problem because all that really 
matters is where the zero-level set lies, and not whether the rest of 
the output far from the zero-level is a "proper" distance map or not. 
If you for some reason need to take this output and make a better 
distance map, your best bet is to use the IsoContourDistanceImageFilter 
and the FastChamferDistanceImage filter to take the narrow-banded 
distance map and create a non-narrow banded Chamfer distance map (which 
is only an approximation to the Euclidian distance, but it's very 
fast!)

Anyhow, what I do when I want to look at a signed distance map is to 
save the actual distance map as a .mhd file of "float" data type. This 
gives a .mhd text file with metadata, and a .raw binary file with the 
actual image data. Then I open this raw file in ImageJ 
(http://rsb.info.nih.gov/ij/ ) with the File > Import > Raw command. 
(Using the dimension information I can read from the .mhd text file.) 
Probably one of the ITK sample applications can read mha/mhd files 
natively, but I like ImageJ. This is a lot more direct for validating 
signed distance map output than saving as binarized images and changing 
the threshold parameters.

Zach Pincus

Department of Biochemistry and Program in Biomedical Informatics
Stanford University School of Medicine


On Jun 27, 2005, at 12:04 AM, Yixun Liu wrote:

> Hi,
> I follow the example using ShapeDectionLevelSetImageFIlter in the 
> ITKSoftwareGuide. The guide says that the output of this filter is a 
> distance map. So I change the parameters of the binarythresholdfilter 
> to validate it. Firstly, I set the lowerThershold -10, upperThreshold  
> 0. Then I change the lowerThreshold to -60, upperThreshold remain 0. 
> However, the segmented result is identical. If the output of the 
> filter is a distance map, the segmented result should be different. I 
> do not know why the results are same. Maybe due to the narrow band? 
> Hope your help!
>  
> Regards,
>  
> Yixun Liu_______________________________________________
> Insight-users mailing list
> Insight-users at itk.org
> http://www.itk.org/mailman/listinfo/insight-users