[Insight-developers] LabelMaps and the Perimeter Estimation

[Bradley Lowekamp]

Gaetan,

I apologize if I came off sounding a bit harsh. I'll see what I can do to try to help with this issue. You have quite a good reputation for contributing very high quality filters, and it's very much appreciated.

I was looking over the code, and it's not clear to me what the definition of perimeter is that you implemented. I am not familiar with Crofton's constant either. Do you have a reference for how you computed this?

I expected the perimeter to be the sum of the "area" of the exterior faces of the labeled pixel. But clearly something else of being computed.


Brad

On Apr 28, 2011, at 4:16 AM, Gaëtan Lehmann wrote:

> 
> Le 27 avr. 11 à 17:52, Bradley Lowekamp a écrit :
> 
>> Gaetan,
>> 
>> Have you seen the following bug:
>> http://www.itk.org/Bug/view.php?id=11934
>> 
>> I encountered this bug myself. It is quite serious because the roundness attribute can become quite erroneous, as I have seen it >2.0 when theoretically it should always be less then 1.0. The miscalculation is causing me to have some trust issues with using this code and relaying on the computed attributes.
>> 
>> In this current state I would not recommend moving the LabelMaps out of ITK review.
>> 
>> Your IJ paper says "More details about perimeter estimation will be published in another article". I only bring up the issue because LabelMaps are quite useful, and would like they to quickly be placed in the reliable category.
> 
> 
> Brad,
> 
> I've seen that problem, but didn't have any chance to fix it.
> I'm not completely sure, but this may be more a problem with the perimeter estimation method than an implementation problem.
> 
> I agree that there is some work to do here, but conclude that the whole label map package is not reliable simply by looking at the lack of precision in perimeter (and thus roundness) estimation doesn't seem very fair. The perimeter estimation is a very small part of it, and other parts are producing a higher precision than what is found currently in ITK.
> 
> Of course some help would be welcome to track that problem, for example by creating tests with expected values.
> 
> Regards,
> 
> Gaëtan
> 
> 
> -- 
> Gaëtan Lehmann
> Biologie du Développement et de la Reproduction
> INRA de Jouy-en-Josas (France)
> tel: +33 1 34 65 29 66    fax: 01 34 65 29 09
> http://voxel.jouy.inra.fr  http://www.itk.org
> http://www.mandriva.org  http://www.bepo.fr
> 

========================================================
Bradley Lowekamp  
Lockheed Martin Contractor for
Office of High Performance Computing and Communications
National Library of Medicine 
blowekamp at mail.nih.gov


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.itk.org/mailman/private/insight-developers/attachments/20110428/2b959562/attachment.htm>