[Insight-users] Geospatial registration

[Luis Ibanez]

Hi Garrett,

Thanks for the detailed description of your project.


>> Any suggestions on good metrics and size of the image to analyze let me 
>> know.
>> 


If you are registering multimodality images, your best
options for metrics are the variants of Mutual Information.
There are about 4 different implementations of Mutual Information
in ITK, you probably want to start with MattesMutualInformation,
that is one of the smoothest.

For a full list of the ImageMetrics available in ITK please look at
http://www.itk.org/Insight/Doxygen/html/group__RegistrationMetrics.html

Note that you can always write your own customized Image Metric and
replace it in the registration framework as you just did with the
transform.


About the computation of derivatives, the size of the perturbations
to be used for computing derivatives by finite differences is something
that must be customized based on the dynamic range of every individual
parameter, as well as the sensitivity of the transforms to those 
particular values.  For example, in a similarity transform, the scale
and rotation parameters are very sensitive to small variations.

This is one of those issues that can only be solved by experimenting
with the values.



    Regards,


       Luis


----------------------
Garrett Potts wrote:

> Hello Luis:
> 
> I am currently testing the registration on a CIB (controlled image 
> block) and a Landsat pan band.  Eventually I would like to do cross 
> sensor registration and eventually batched bundle adjustments.  I have a 
> 5 meter and a 10 meter CIB file and the landsat completely overlaps the 
> CIB data set.  I currently just hand picked a point that looks like it 
> should correlate good.  The Landsat models are around a few pixels off.  
> Within our software we have parameters that can be tweaked in order to 
> adjust and remove the errors.  I figured this would be nice to bridge 
> since Insight package works well with adjustable parameters.
> 
> Currently I have the following:
> 
> 1.  Derived a transform class that bridges to our adjustable 
> parameters.  Since I can't calculate the partials explicitly I 
> numerically calculate it by varying the adjustable a small amount and 
> see how x and y change and set the Jacobian.  I am not sure what the 
> ideal perturbation of each parameter should be to calculate an accurate 
> partial.
> 
> Every time the SetParameter is called I adjust our internal transform 
> found in our internal image processing chain.
> 
> 2.  Derived an interpolator.  Just returns a pixel value in the floating 
> image chain.
> 
> Note:  Both images are resampled and projected to a Map view and all 
> analysis is done there.  Given that, I currently don't transform any 
> points coming into the transform.
> 
> 
>> You may want to look closer at the size and location of the
>> image regions that you want to be taken into account while
>> the image metric is computed.  In particular that will
>> affect your selection of the FixedImageRegion() and eventually
>> the use of an Image Mask.
> 
> 
> I have no clew what the best dimension is to analyze and am not sure 
> what metric works best.  Currently, I think the Mutual information 
> metric is working well.  I am still having problems with parameter 
> adjustments and getting the parameters to settle down to an accurate 
> adjustment.
> 
> Any suggestions on good metrics and size of the image to analyze let me 
> know.
> 
> 
> Ultimately I would like to automatically find the correlation points on 
> the image and then take the minimal fit of all points and use the result 
> as the final adjustment.
> 
> 
> Thank you for the help and
> 
> Take care
> 
> 
> Garrett
> 
> _______________________________________________
> Insight-users mailing list
> Insight-users at itk.org
> http://www.itk.org/mailman/listinfo/insight-users
> 
>