[Insight-users] Re: metrics and image masks

[Luis Ibanez]

Hi Stefan,

That's a good observation.

What happens is that unfortunately the toolkit is not fully
consistent on the use of rounding versus truncation when
converting continuous index to index.

Please add this as a bug entry to the phpBugTracker. It is
something that will require to review many classes across
the toolkit in order to make them consistent.

We probably should do that by providing a helper class or
a helper macro that will do the appropriate (round or truncate)
consistently everywhere.


Let us know if you encounter any problem for creating
a bug entry for this issue.


    Thanks


       Luis


--------------------
Stefan Klein wrote:
> Hi,
> 
> Related to this subject I have two questions:
> 
> - ImageSpatialObject.txx line 177: why is the index computed by a cast, 
> and not using a 'round'? the ITK software guide says the value of the 
> closest pixels is returned, so nearestneighbour interpolation. A round 
> operation sounds then more logical to me. The same applies to line 62 of 
> ImageMaskSpatialObject by the way.
> 
> - ImageSpatialObject.txx line 219: why is "size[i]" used and not 
> "size[i] - 1" ? Isn't this the underlying cause of the troubles 
> Christoph has? Alternatively,
>         pointLow[i] = -0.5 + epsilon;
>         pointHigh[i] = size[i] - 0.5 - epsilon;
> may make more sense, if you consider Figure 4.1 in the software guide. 
> But this is only allowed if you use a round operation in line 177.
> 
> Best regards,
> Stefan.
> 
> 
> At 11:53 21/08/06, Luis Ibanez wrote:
> 
>> Hi Christoph,
>>
>> Thanks for posting all the details of the traceback.
>>
>> This illustrates that there is a bug in the ImageMaskSpatialObject.
>>
>> The class is using the index without first verifying if it is inside
>> the BufferedRegion or not.
>>
>> This issue has been entered as bug #3645 in the phpBugTracker
>> http://public.kitware.com/Bug/bug.php?op=show&bugid=3645&pos=0 
>> <http://public.kitware.com/Bug/bug.php?op=show&bugid=3645&pos=0>
>>
>> A fix has been committed to CVS
>> http://www.itk.org/cgi-bin/viewcvs.cgi/Code/SpatialObject/itkImageMaskSpatialObject.txx?root=Insight&sortby=date&r2=1.10&r1=1.9 
>> <http://www.itk.org/cgi-bin/viewcvs.cgi/Code/SpatialObject/itkImageMaskSpatialObject.txx?root=Insight&sortby=date&r2=1.10&r1=1.9>
>>
>> Please update your CVS checkout of ITK and let us know if you
>> continue experiencing any problems.
>>
>>
>>     Thanks
>>
>>
>>        Luis
>>
>>
>>
>> ------------------------------
>> Christoph Niedermayr wrote:
>>
>>> Am Donnerstag, den 17.08.2006, 15:13 +0200 schrieb Christoph Niedermayr:
>>>
>>>> Am Donnerstag, den 17.08.2006, 08:39 -0400 schrieb Luis Ibanez:
>>>
>>>
>>>>> What version of ITK are you using ?
>>>>
>>>>
>>>> 2.6
>>>>
>>>>
>>>>> There are no calls to GetPixel() inside the file
>>>>>    Insight/Code/Algorithms/
>>>>>      itkMattesMutualInformationImageToImageMetric.txx
>>>>> What the metric calls is the method "IsInside" in lines
>>>>> 457 and 526.
>>>>
>>>>
>>>> you're right, i got that confused. But "IsInside" then calls "GetPixel",
>>>> if the sample is considered inside the SpacialObject, in my case with an
>>>> invald index.
>>>
>>>
>>> Hi Luis!
>>> here's whats happening exactly, values are taken from gdb:
>>> the image i use as mask has the following properties:
>>> origin = [-150, -113, -231]
>>> index  = [0, 0, 0]
>>> size   = [60, 42, 60]
>>> spacing= [5, 5, 5]
>>> the point that is given to "IsInside" has the coordinates
>>> [30, -88, 69]
>>> and is considered inside (because the upper z bound of the spacialobject
>>> is 69, and only values exceeding the bound are considered outside)
>>> z-distance between origin and the point is 300.
>>> z-index is then calculated as 60 (invalid), causing SIGSEV when calling
>>> "GetPixel"
>>> below you find a code sample which causes this error. of course, SIGSEV
>>> only occurs if the accessed memory location is indeed off limits to the
>>> program. eg. using a 2D image of size 10x10 queries the image buffer at
>>> 110, but did not cause SIGSEV..
>>> best regards,
>>> chris
>>> _____
>>> #include "itkImage.h"
>>> #include "itkImageMaskSpatialObject.h"
>>>   // create 123x123x123 image
>>>   typedef itk::Image<unsigned char,3> ImageType;
>>>   ImageType::Pointer img = ImageType::New();
>>>   ImageType::RegionType region;
>>>   ImageType::SizeType size;
>>>   ImageType::IndexType index;
>>>   ImageType::PointType origin;
>>>   ImageType::SpacingType spacing;
>>>   size.Fill(123);
>>>   index.Fill(0);
>>>   origin.Fill(0);
>>>   spacing.Fill(1);
>>>   region.SetSize(size);
>>>   region.SetIndex(index);
>>>   img->SetRegions(region);
>>>   img->SetOrigin(origin);
>>>   img->SetSpacing(spacing);
>>>   img->Allocate();
>>>   // create mask from image
>>>   itk::ImageMaskSpatialObject<3>::Pointer mask =
>>> itk::ImageMaskSpatialObject<3>::New();
>>>   mask->SetImage(img);
>>>   // check if borderpoint is inside
>>>   ImageType::PointType testPoint;
>>>   testPoint.Fill(123);
>>>   mask->IsInside(testPoint);
>>> _______________________________________________
>>> Insight-users mailing list
>>> Insight-users at itk.org
>>> http://www.itk.org/mailman/listinfo/insight-users
>>
>>
>>
>> _______________________________________________
>> Insight-users mailing list
>> Insight-users at itk.org
>> http://www.itk.org/mailman/listinfo/insight-users
> 
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> Insight-users mailing list
> Insight-users at itk.org
> http://www.itk.org/mailman/listinfo/insight-users