[Insight-users] Kd-tree implementation in ITK is buggy and unreliable

[Luis Ibanez]

Hi Ali,

Thanks for the additinonal information.

      Yeap, the algorithm should work regardless
      of the bucket size. That one was a false alarm.


I'm tracking now the cases where itkKdTreeTest2 segfaults
with a bucket size == 2.  Hopefully this will illuminate
at least part of the problem.


   Thanks


      Luis


---------------
Ali - wrote:
> Luis,
> 
> I used a bucket size of 16 with about 100 particles. Like you mentioned, the algorithm must work fine independent of these parameters.
> 
> 
> -Ali
> 
> 
>>
>>Hi Ali,
>>
>>
>>Could you please report
>>
>>
>>   1) How many points (measurement vectors) are you
>>      currently feeding into the KdTreeGenerator ?
>>
>>and
>>
>>
>>   2) What value of "BucketSize" did you use ?
>>
>>
>>
>>The problem reported in the bug
>>http://public.kitware.com/Bug/view.php?id=5082
>>
>>Seems to be simply the result of a poor choice
>>on the value of the bucket size parameter.
>>For the test case reported in that bug, increasing
>>the value from 16 to 30 solves the problem.
>>
>>We are wondering if the problem that you are
>>experiencing is also the result of poor parameter
>>setting.
>>
>>
>>    Please let us know,
>>
>>
>>       Thanks
>>
>>
>>
>>           Luis
>>
>>
>>
>>------------
>>Ali - wrote:
>>
>>>Hi,
>>>
>>>A quick search in the mailing list shows that, during the past half a decade, it has been reported many times that the kd-tree classes in ITK are buggy. Writing some library based on ITK, it took me a few days to find out that the bug in my library is actually introduced by the kd-tree implementation in ITK which FINDS THE WRONG NEAREST NEIGHBOUR.
>>>
>>>I have no idea, against many warnings from the users, why the bug has not been resolved yet. In the case it is difficult to address the bug, one option is to wrap many other existing implementations, personally I switched to libkdtree++.
>>>_________________________________________________________________
>>>100’s of prizes to be won at BigSnapSearch.com 
>>>http://www.bigsnapsearch.com
>>>_______________________________________________
>>>Insight-users mailing list
>>>Insight-users at itk.org
>>>http://www.itk.org/mailman/listinfo/insight-users
>>>
> 
> 
> _________________________________________________________________
> 100’s of prizes to be won at BigSnapSearch.com 
> http://www.bigsnapsearch.com