[Insight-users] itk::NormalizedCorrelationImageToImageMetric

[Luis Ibanez]

Hi Kathy,

The reason why the resample filter only produces one image is that
only the "Moving" image needs to be transformed. The transformation
is such that the moving image is presented in the coordinate system
of the "Fixed" image.

If you want to evaluate the quality of the superposition, what you
can do is to compute the differences between the Fixed image and
the Resampled Moving image. Note that you should only visually look
at the region where the two images get overlapped.

You will find many examples on how to compare the Fixed image to the
Resampled Moving image in the ITK Software Guide


     http://www.itk.org/ItkSoftwareGuide.pdf


in the "Image Registration" chapter.

The source code of the examples illustrated in the ITK Software Guide
can be found in the directory:


        Insight/Examples/Registration


Please read the "Image Registration" chapter of the software guide,
it will help you understand the ITK registration framework and this
will save you a lot of time and effort down the line.


Regards,


    Luis



================================
smythek at u.washington.edu wrote:
> Hi Jim,
> 
> I tried the registration example with the 
> NormalizedCorrelationImageToImageMetric instead of the 
> MeanSquareImageToImageMetric, and it gave me an optimal transform after 
> one iteration.  This could be correct, but the resample filter only 
> writes out the second image.  I want it to write out both to the same 
> file so I can see the new (aligned) image. I don't understand why the 
> resample filter doesn't write out both images.  How do i get it to do so?
> 
> Thank you,
> 
> Kathy
> 
> 
> On Sun, 7 May 2006, Jim Miller wrote:
> 
>> The NormalizedCorrelationImageToImageMetric is just a metric.  You 
>> provide
>> it with two images and a transform.  The registration framework can be 
>> used
>> to adjust the parameters of the transformation to identify the optimal
>> transformation to align the images.  In this context, the transformation
>> relates the physical space (mm, ft, etc.) of the two images not their
>> indices.
>>
>> You have a choice of transformations: rigid, affine, deformable to define
>> the relationship between the two image coordinate frames.  You also 
>> have a
>> choice of optimization methods: gradient descent, lbfgs, amoeba, etc.
>>
>> You also have the freedom to employ the metric yourself by providing a
>> series of transformations.  For instance, you could uniformly sample a 
>> set
>> of translations to identify which translation gives the best normalized
>> correlation value.
>>
>> Jim
>>
>> On 5/6/06, k s <smythek at u.washington.edu> wrote:
>>
>>>
>>> Hi ITK,
>>>
>>> I am a new ITK user and I wanted to stitch together a couple hundred of
>>> tiff images using correlation.  My question is: does this normalized
>>> correlation filter return the indexes of the best correlation between 
>>> two
>>> images in addition to the value?
>>>
>>> Thanks,
>>>
>>> Kathy
>>>
>>> _______________________________________________
>>> 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
> 
>