[Insight-users] Bug in GeodesicActiveContourImageFilter????????

[Oliver Gloger]

Hello again,

is there anyone who used the GeodesicActiveContourImageFilter??? I read the softwareguide, the examples and got a little bit involved in Level Sets (/ geodesic active contours) but do not manage to achieve fine results with this filter!!!! The segments are growing too large and it seems that they do not notice the image gradients. 
1.)	In the SoftwareGuide is written: 
sigmoid->SetOutputMinimum(  0.0  );
   sigmoid->SetOutputMaximum(  1.0  );
	and in the figures of the softwareguide (e.g. figure 9.19) you can notice that the outputimage of the gradientmagnitude has higher values at edges(white), but after using the sigmoidfilter it has lower values at the edges (black)!!! 
	This is only achievable by choosing the opposite values:
sigmoid->SetOutputMinimum(  1.0  );
   sigmoid->SetOutputMaximum(  0.0  );
Only by this way the Sigmoidfilter produces corresponding images. Is this true?
2.)	My Inputimage of the gradientmagnitude has grayvalues ranging from 0..255 and i set alpha=255 and beta=127, but the Geodesic...Filter doesn’t notice the edges resp. gradients and delivers too large grown segments form my initial segments. Only by setting alpha=1 and beta=0.5 I get some segments which I could regard as “correct evolved segements”. I found it strange! Is there anyone who could explain me this?
Could there be a bug i this filter (the Laplacian Level Set and Canny Edge Level Set Segementation Filters deliver correct results!!!!)

How to tell the filter that he schould stop at the higher gradients?

Best regards, 

Olli


-- 
"Feel free" - 5 GB Mailbox, 50 FreeSMS/Monat ...
Jetzt GMX ProMail testen: www.gmx.net/de/go/mailfooter/promail-out