[Insight-users] Huge performance drop of BinaryErodeImageFilter

[nicolas.savoire at gmail.com]

Hi all,

I recently upgraded to ITK 3.2.0 and noticed a huge slow down of my 
application. I narrowed it down to itk::BinaryErodeImageFilter which seems to 
have changed since ITK 3.0.0. 
With ITK 3.0.0, the erosion on a 896x1410 unsigned int image with a ball 
structuring element element took 100ms, where as it now takes more than 5 
minutes for the same operation  !!!
Both versions are compiled with gcc 4.1.2 and relwithdebinfo optimization 
flags.
I am a little confused since the itk journal submission referring to this 
filter (http://hdl.handle.net/1926/173) claims that the new filter 
outperforms the old one. Is the structuring element I am using too big ?
Has any of you experienced the same behavior ?

best regards,
nicolas