Open mileslucas opened 1 year ago
@mileslucas have you attempted to add this feature to the code? What is different about the 3D case?
No I have not attempted this- feel free to try yourself if you'd like. I believe it's just an extension of the Fourier transforms into more dimensions than two.
It looks like the only thing that needs to change is the upsampled_dft function; in particular, we need some kind of equivalent of _data = kernel * _data'
for data with 3 dims (adjoint is only defined in 2 dims). It would be a tensor product I guess? Is there something like that in Julia?
nevermind, I figured it out. I'll make a PR soon
it would be so noce to have the multi-dimensional functionality back. It worked very nicely on 3d MRI data. @RpfR2000 Any chance you might like to share your solution ? In any case, thanks for this nice package.
The original Julia version of this package had support for multi-dimensional registration (e.g., registering a 3-D cube instead of a 2-D image). This was removed for simpler code, but should be added back as long as it doesn't negatively affect the performance of the 2-D registration.