[Insight-users] Heads up on HessianRecursiveGaussian bug fix

[Luca Antiga]

Hi Rahul,
  the fixed itkHessianRecursiveGaussianImageFilter is in the ITK CVS  
repository (Insight/Code/BasicFilters) and it has been backported  
today to the 3.18 branch by Luis. The test is in Insight/Testing/Code/ 
BasicFilters.
To test your program you'll have to checkout ITK from there and build  
it, and then compile your program against it.

As a cheap alternative for testing purposes, you can download the  
incriminated file (http://public.kitware.com/cgi-bin/viewcvs.cgi/Code/BasicFilters/itkHessianRecursiveGaussianImageFilter.txx?revision=1.10&root=Insight&view=markup 
), put it in your own source tree at Insight/Code/BasicFilters (i.e.  
replacing your own version of the file) and compile your program.

Hope this helps


Luca



On Apr 28, 2010, at 12:09 PM, Rahul P.Kumar wrote:

> Hi Luca,
>
> May I know where I can find the newly fixed  
> itkHessianRecursiveGaussianImageFilter and  
> tkHessianRecursiveGaussianImageFilterScaleSpaceTest for testing my  
> program?
>
> Regards,
> Rahul
>
> From: Luca Antiga <luca.antiga at gmail.com>
> To: itk Users <insight-users at itk.org>
> Sent: Wed, April 28, 2010 3:25:41 PM
> Subject: [Insight-users] Heads up on HessianRecursiveGaussian bug fix
>
> Hi all,
>  this is a heads up message on a bug that Oleksandr Dzyubak has  
> found in itkHessianRecursiveGaussianImageFilter
> (bug 0008990 in Mantis, http://public.kitware.com/Bug/view.php?id=10599) 
>  and for which I recently submitted a fix.
>
> The bug affected the behavior of the filter when  
> NormalizeAcrossScale is set to true with dimensions > 2, and it has  
> effects on multi-scale vesselness filters or multi-scale shape- 
> detection filters in general (e.g. those using  
> itk::MultiScaleHessianBasedMeasureImageFilter), and vessel-enhancing  
> diffusion filters.
>
> The bug caused incorrect normalization of the Hessian across scale- 
> space, and, for instance, lead to a poor performance of vesselness  
> filters when spanning several scales (vessels at smaller scales were  
> washed out compared to how they appear after the fix).
>
> The fix has been committed after 3.18 was released, hence it will be  
> in for ITK > 3.18. There is now a test  
> (itkHessianRecursiveGaussianImageFilterScaleSpaceTest) that tests  
> the scale-space behavior of the class quantitatively.
>
> So, for whoever was using multi-scale Hessian-based filters in 3D  
> and higher, please rerun your tests, since there's a very good  
> chance (100% ?) that the results will have changed (most likely for  
> the better).
>
> I'm here for any clarification. Again, kudos to Oleksandr Dzyubak  
> who highlighted the misbehavior through lots of testing.
>
> Best regards
>
>
> Luca
>
>
>
> --
> Luca Antiga, PhD
> Principal Scientist, OROBIX Srl
> via L.A. Muratori 3, 24123 Bergamo, Italy
>
> web: www.orobix.com
> mobile: +39.347.43.16.596
>
> "This message originates from OROBIX Srl and its contents and  
> attachments are privileged and confidential and are intended only  
> for the individual(s) or entity(ies) named above. This message  
> should not be forwarded, distributed or disclosed. Any unauthorized  
> use, dissemination and duplication is strictly prohibited and may be  
> unlawful. All personal messages express views solely of the sender,  
> which are not to be attributed to OROBIX Srl, and may not be copied  
> or distributed without this disclaimer. If you are not the intended  
> recipient or received this message in error, please delete this  
> message and notify the sender by reply e-mail. Opinions, conclusions  
> and other information in this message that do not relate to the  
> official business of OROBIX Srl shall be understood as neither given  
> nor endorsed by it."
>
>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.itk.org/pipermail/insight-users/attachments/20100428/be7524f2/attachment-0001.htm>