[Insight-users] Mutual Information Error

[Jordan Soet]

Well, I thought I had fixed my problem, but it looks like I didn't..


After my last e-mail where I mentioned that I got the error message even 
if I set the fixed and moving image to the same image I started looking 
to see if maybe it was something to do with the random number generator, 
and through looking through the mailing list I realized that I hadn't 
initialized the seed. So, I added metric->ReinitializeSeed() and it 
seemed to fix my problem, I could get ImageRegistration11 working on 3D 
images.

However, ImageRegistration11 only uses a translation transform, so I 
changed it to an affine transform and the problem came back. I again got 
the message "Too many samples map outside moving image buffer: 0 / 
10000". I then tried changing it back to 2D while still using an affine 
transform and it worked fine. I then tried using 
MultiResImageRegistration2 and changing it to 3D, since it already uses 
an affine transform, and would remove the possibility that I was making 
errors with the transform, but again, it worked on 2D images but when I 
moved to 3D images it gave me the exception. Also, I tried these all 
using the same image for both inputs, as well.

In addition, I also tried testing my random number generator by building 
itkImageRandomIteratorTest, as suggested in another e-mail, but I got 
the following error:

Building executable /home/Jordan/Test/itkImageRandomIteratorTest.exe...
/usr/lib/gcc-lib/i686-pc-cygwin/3.3.3/../../../libcygwin.a(libcmain.o)(.text+0xb
3): undefined reference to `_WinMain at 16'
collect2: ld returned 1 exit status
make: *** [/home/Jordan/Test/itkImageRandomIteratorTest.exe] Error 1


Thanks in advance for any advice,
Jordan





Luis Ibanez wrote:

>
> Hi Jordan,
>
>
> This message is typically produced when the Transform is such that
> the Moving image is mapped outside the extent of the Fixed image.
>
> The lack of overlap between the Fixed image and the mapped Moving
> image makes impossible to compute the Metric, and therefore an
> Exception is thrown.
>
>
> Please post to the list the characteristics of your two images.
>
>      - Number of pixels along each dimension
>      - Pixel spacing
>      - Origin
>
>
>
> You can get spacing and origing very easily by just invoking
>
>
>      fixedImageReader->GetOutput()->Print( std::cout );
>      movingImageReader->GetOutput()->Print( std::cout );
>
>    [after having invoked the Update() methods of the readers]
>
>
>
>
> It is very likely that the origin of your images is such that
> initializing the Transform as an Identity is not appropriate.
>
>
>
> In that case, you may want to use the Transform Initializer, as
> illustrated in the ITK Software Guide
>
>
>          http://www.itk.org/ItkSoftwareGuide.pdf
>
>
> Chapter 8,
>
>
>
>    Regards,
>
>
>
>        Luis
>
>
>
> --------------------------
> Jordan Soet wrote:
>
>> I'm working the mutual information metric on 3D ultrasound images, 
>> and as a quick test I tried using ImageRegistration11 on it with the 
>> dimension changed to 3 (and the parameters for the transform changed 
>> as well). However, when I try and run it I get the following error:
>>
>> $ ./ImageRegistration11.exe I0000005.hdr I0000006.hdr
>> ExceptionObject caught !
>>
>> itk::ExceptionObject (0xa05c370)
>> Location: "Unknown"
>> File: 
>> /home/Jordan/InsightToolkit-2.0.1/Code/Algorithms/itkMattesMutualInformationImageToImageMetric.txx 
>>
>> Line: 623
>> Description: itk::ERROR: 
>> MattesMutualInformationImageToImageMetric(0xa0528a0): Too many 
>> samples map outside moving image buffer: 0 / 10000
>>
>> I'd guess that it's something to do with the regions, however, the 
>> images are the same size, and I didn't change anything from the 
>> example, so I don't know what the matter is... Does anyone know what 
>> the problem is?
>>
>>
>> Thanks,
>>
>> Jordan
>> _______________________________________________
>> Insight-users mailing list
>> Insight-users at itk.org
>> http://www.itk.org/mailman/listinfo/insight-users
>>
>>
>
>
>