[Insight-developers] some comments about ImageBoundaryFacesCalculator changes

[Bradley Lowekamp]

Bill,

I discovered this myself a couple weeks ago (posted on this list). As it is causing issues with my threading of a method in the AnisotropicDiffusion filters. 

I would consider this behavior a bug myself. But there is a note in the txx file about this behavior, but there is no doxygen documentation. My assumption was that faces where not intersecting. 

Brad

On Nov 17, 2010, at 11:37 AM, Bill Lorensen wrote:

> Wanlin,
> 
> This original behaviour may be a bug. I don't think there should be
> duplicate boundary regions. Maybe someone more familiar with the
> calculator can comment.
> 
> If it turn out to be a bug, then I don't think we'll need to turn the
> bug on/off.
> 
> Bill
> 
> On Wed, Nov 17, 2010 at 9:18 AM, wanlin <wanlinzhu at gmail.com> wrote:
>>  Hi, David and Bill. Thank for your suggestions. I will add tests, comments
>> and change variable names etc as you have suggests later. I am sorry i don't
>> know how to add comments to gerrit. Following are some explanations.
>> 
>> The ImageBoundaryFacesCalculator will produce duplicate boundary regions at
>> corner.
>> For example, for requestedregion size = 5x5 image with radius = 1, The
>> previous version of ImageBoundaryFacesCalculator will give
>> Number of pixels : 9
>> ImageRegion (0x18ad830)
>>   Dimension: 2
>>   Index: [1, 1]
>>   Size: [3, 3]
>> 
>> Number of pixels : 5
>> ImageRegion (0x18ad870)
>>   Dimension: 2
>>   Index: [0, 0]
>>   Size: [1, 5]
>> 
>> Number of pixels : 5
>> ImageRegion (0x18ad8b0)
>>   Dimension: 2
>>   Index: [4, 0]
>>   Size: [1, 5]
>> 
>> Number of pixels : 5
>> ImageRegion (0x18ad8f0)
>>   Dimension: 2
>>   Index: [0, 0]
>>   Size: [5, 1]
>> 
>> Number of pixels : 5
>> ImageRegion (0x18ad930)
>>   Dimension: 2
>>   Index: [0, 4]
>>   Size: [5, 1]
>> 
>> The updated version takes into account the issue and produce non duplicate
>> regions.
>> 
>> Number of pixels :9
>> ImageRegion (0xb8b830)
>>   Dimension: 2
>>   Index: [1, 1]
>>   Size: [3, 3]
>> 
>> Number of pixels :5
>> ImageRegion (0xb8b870)
>>   Dimension: 2
>>   Index: [0, 0]
>>   Size: [1, 5]
>> 
>> Number of pixels :5
>> ImageRegion (0xb8b8b0)
>>   Dimension: 2
>>   Index: [4, 0]
>>   Size: [1, 5]
>> 
>> Number of pixels :3
>> ImageRegion (0xb8b8f0)
>>   Dimension: 2
>>   Index: [1, 0]
>>   Size: [3, 1]
>> 
>> Number of pixels :3
>> ImageRegion (0xb8b930)
>>   Dimension: 2
>>   Index: [1, 4]
>>   Size: [3, 1]
>> 
>> To keep backward compatibility, the default behavior of
>> ImageBoundaryFacesCalculator has been set the same as previous version.
>> With default option (m_DuplicateOverlap == true, same as previous version),
>> all tests(1822) passed.
>> 
>> If the default option  m_DuplicateOverlap == false, 3 tests failed.
>> 
>> 1. 159 - itkScalarImageToCooccurrenceListSampleFilterTest (Failed)
>> In itkScalarImageToCooccurrenceListSampleFilterTest
>> The baseline value of baselineVectorList has duplication regions. Namely
>> they are using the default behavior (m_DuplicateOverlap == true)
>> 
>> 2. 1811 - GeodesicActiveContourShapePriorCallosumTest (Failed)
>> The errors could be traced back to
>> itk::ScalarAnisotropicDiffusionFunction<Image>::CalculateAverageGradientMagnitudeSquared(TImage
>> *ip)
>> When duplicate regions exist, the gradient magnitudes of part of boundary
>> regions have been counted twice or three times depends on image dimension
>> and locations . I think it is a bug, in the case, m_DuplicateOverlap should
>> be set false to avoid duplication.
>> 
>> 3. 1817 - WatershedSegmentation1Test1 (Failed)
>> The error could be traced back to itk::VectorAnisotropicDiffusionFunction<
>> TImage >::
>> CalculateAverageGradientMagnitudeSquared(TImage *ip) The same reason as 2.
>> 
>> 
>> 
> _______________________________________________
> Powered by www.kitware.com
> 
> Visit other Kitware open-source projects at
> http://www.kitware.com/opensource/opensource.html
> 
> Kitware offers ITK Training Courses, for more information visit:
> http://kitware.com/products/protraining.html
> 
> Please keep messages on-topic and check the ITK FAQ at:
> http://www.itk.org/Wiki/ITK_FAQ
> 
> Follow this link to subscribe/unsubscribe:
> http://www.itk.org/mailman/listinfo/insight-developers

========================================================
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/20101117/41db79a9/attachment.htm>