[Insight-developers] itkHistogram.hxx - values equal to max bin are missed

[Matt McCormick]

Hi Matt,

As a side note, there is a migration guide for updating the Statistics
framework here [1], and there is no mention of this change of behavior
there that I can see.


Using 'git blame', it appears that this bug was introduced when
migrating the statistics refactoring to default in this commit[2].

I created an issue to track this bug [3].  If you are interested in
submitting a patch, instructions can be found here [4].

Thanks,
Matt

[1]http://www.itk.org/Wiki/Proposals:Refactoring_Statistics_Framework_2007_Migration_Users_Guide
[2] http://itk.org/gitweb?p=ITK.git;a=commit;h=c6d83112
[3] https://issues.itk.org/jira/browse/ITK-3206
[4] http://insightsoftwareconsortium.github.io/ITKBarCamp-doc/CommunitySoftwareProcess/SubmitAPatchToGerrit/index.html

On Thu, Aug 29, 2013 at 12:13 PM, Clarkson, Matt <m.clarkson at ucl.ac.uk> wrote:
> Hi there,
>
> I am upgrading from ITK 3.20 to 4.3.2 and noticed that in 4.3.2 a histogram
> with 64 bins (0-63), spanned by the image values 0-255 mapped the value 255
> to bin 64, which was subsequently missed. This worked in 3.20, resolving to
> bin 63.
>
> Comparing the code of itkHistogram.txx in 3.20, I see in method:
> ::GetIndex(const MeasurementVectorType & measurement,IndexType & index )
> const
>
> code that does:
>
>     if (tempMeasurement >= m_Max[dim][end])
>       {
>       // one of measurement is below the minimum
>       // its ok if we extend the bins to infinity.. not ok if we don't
>       //Need to include the last endpoint in the last bin.
>       if(!m_ClipBinsAtEnds || tempMeasurement ==  m_Max[dim][end])
>         {
>         index[dim] = (long) m_Size[dim]-1;
>         continue;
>         }
>
> whereas in version 4.3.2 I see:
>
>
>     if ( tempMeasurement >= m_Max[dim][end] )
>
>       {
>
>       // one of measurement is above the maximum
>
>       // its ok if we extend the bins to infinity.. not ok if we don't
>
>       if ( !m_ClipBinsAtEnds )
>
>         {
>
>         index[dim] = (IndexValueType)m_Size[dim] - 1;
>
>         continue;
>
>
>
> So, it looks to me like version 4.3.2 is missing the condition where if a
> value is exactly equal to the upper bound then it is included.  Then I went
> to JIRA and found:
> https://issues.itk.org/jira/browse/ITK-2266
>
> which suggests that this was a bug at one point, but was fixed in 3.20.
>
> So, did I miss a link somewhere? Can someone advise me please?
> Has the spec changed such that the upper bound is no longer meant to be
> inclusive in version 4+?
>
> Thanks
>
> Matt
>
> ------------------------------------------------------
> Matt Clarkson Ph.D.
> CMIC Software Manager
> Senior Research Associate
> m.clarkson at ucl.ac.uk
> Skype: drmattclarkson
>
> Centre For Medical Image Computing
> http://cmic.cs.ucl.ac.uk/staff/matt_clarkson/
> Tel: 020 7679 0257
> Fax: 020 7679 0255
> Room: 2.21 Malet Place Engineering Building
>
> Dementia Research Centre
> http://dementia.ion.ucl.ac.uk/
> Fax: 020 7676 2066
> ------------------------------------------------------
>
>
>
>
>
>
>
>
>
> _______________________________________________
> 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.php
>
> 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
>