spatial correlation of 3d matrices

Hi, I'm trying to compute the similarity of two 3d images. Is there any way to get the spatial correlation of these two images? Many thanks!

Answer 1

[Matt J]

How about

ifftn(fftn(Image1).*conj(fftn(Image2)),'symmetric')

with zero-padding if appropriate.

[Liang]

Thanks, Matt. I've been told that corrcoef can also be used to compute the spatial correlation of 3d matrices. I think this may be a simple way, although it may not give out the similarity of images precisely. Could you give me some suggestion on the difference between your algorithm and corrcoef? Sorry for the stupid question...

[Matt J]

What I've proposed is basically a 3D version of XCORR2, implemented with FFTs.

[efe ozkaya]

Hi Matt, is there a need to use fftshift in the code you provided? Should I be using it like this instead?

fftshift(ifftn(ifftshift(fftshift(fftn(Image1)).*conj(fftshift(fftn(Image2)),'symmetric'));

Also, what would you suggest for observing correlation map in 3D?

One more thing, let's say that I have a lot of zeros in the 3D volume in order to avoid their contribution to correlation what can I do?

[Matt J]

Hi efe,

fftshifting should not be necessary.

Viewing the 3D correlation map - there are a variety of 3D image viewers offered on the File Exchange. You might check there.

Avoiding contributions of zero. Crop them out? I think you can also do effectively the same thing by using the fft syntax

Y = fftn(X,sz)

were sz covers the part of the image you want to include.

[Naomi Krauzig]

Hi Matt in case the question is still open there is a 3d version available, it's called spatialcorr3 https://it.mathworks.com/matlabcentral/fileexchange/65640-spatialcorr3-varargin from Ankur Kumar.

All the best,

Naomi