[Insight-users] registration of multimodal images and optimisation surface

[Severino]

I am having trouble to register two small images coming from the remote sensing area, one a radar image and the other one from an optical sensor. Mutual information hould work for this case as it does in medicine, to test it I used an application written by the ORFEO project from CNES. But trying with a large set of different parameters for the optimization (gradient descent) and for the measure (Viola and Wells or Mattes).
I have therefore used the small program of Lydia Ng to observe the behaviour of the function to be optimized. But the results are quite astonishing, as I can not see a global maximum but a plateau like maximum. If this is the real shape, it is no wonder that I can not register the function, as there are a compact set of maxima. I suppose this strange shape comes froom the perameter settings, but I just used the default values that Lydia put in her code, and any variation showed the same shape.
Is anybody aware of settings I could try to see if this shape changes and produces a smooth surface with a well identified maximum?
regards 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.itk.org/pipermail/insight-users/attachments/20080418/03e4968d/attachment.htm>