[Elastix] input images gray level
s.klein at erasmusmc.nl
Thu Jul 25 22:40:16 CEST 2013
Elastix does not do intensity rescaling, but if you use normalized mutual information (or mutual information, or normalized correlation) that is done implicitly:
a) in case of MI and NMI: the joint histogram bins are spaced such that they cover the entire range of the image; so you could say it implicitly does a rescaling.
b) the normalized correlation is defined exactly such that it is invariant to such global scalings.
Regarding denoising, elastix uses a multiresolution approach, in which it starts with blurred images, so in a way that's denoising.
On 25-7-2013 22:28, Li, Wen wrote:
I have CT images converted from original dicom files, so it has 16bit gray levels. I used them as input images and got decent results by using B-spline with normalized mutual information (NMI).
I wonder if Elastix does any sort of intensity rescaling or denoising internal before calculates NMI? I can't find it in the source code.
Elastix mailing list
Elastix at bigr.nl<mailto:Elastix at bigr.nl>
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Elastix