[Insight-users] Compute the overlapping volume of two images (by SimilarityIndexImageFilter)

[Michael Schildt]

Hello Xabier!

I try to explain it with other words. I acquired two anatomical data 
sets within one MRI session (3 slices, 256x256 inplane, 4 mm slices 
thickness) with identical parameter. They contain exactly the same 
volume of the brain. Within ITK I do a rigid registration to a third 
high resolution 3D volume (from another scanner) for each of the two 
data sets. (All 3 data sets are T1 weighted.)

I expect for a perfect registration, that the slice data sets are 
positioned exactly at the same position and orientation. To test this, i 
would like to compute this overlap. Until now, i do this by computing 
the euklidian distance of the image centers and the squared differences 
in rotation angles. But the overlap includes both errors and scales from 
0.0 to 1.0.

Does it help clearifying my issue?

Greetings,
    Michael

Xabier Artaechevarria Artieda schrieb:
> Hi Michael,
> well, this filter is thought to evaluate differente methods to segment 
> the "same object in the same image". Usually, you want to see how 
> similar is the volume segmented by method A to the gold standard. In 
> that context, the inputs to the filter are 2 binary images of the same 
> size. The regions of interest (on which the filter works) are also the 
> same.
> I must say I do not really understand what you are looking for. But 
> from your implementation idea, wouldn't it be enough to resize one of 
> the images to fit the size of the other? If you explain your problem 
> in more detail I might give you some other idea.
>
> Grüsse,
> Xabi
>