I'm replying here because the subject is more relevant and the thread is more recent, but I'll quote Emmanuelle from the older thread "Segmentation algorithms".
On Friday, June 15, 2012 4:28:07 PM UTC-4, Emmanuelle Gouillart wrote:
I could, however, write a tutorial on how to deal with the processing of
3-D images with numpy/skimage/mayavi...
I would be very interested in such a tutorial. :)
Since I'm using skimage for processing mostly 3-D images (from X-ray
tomography), I'm strongly biased towards 3-D support. But in fact, it's
only worth implementing an algorithm for 3-D images if it's fast enough
and does not have too high a memory load, since the data are much
larger (nobody wants to wait one day for the segmentation of a
1000*1000*1000 image... whereas one minute for a 1000x1000 image might
still be decent enough in 2-D).
Do you have thoughts on 3D image cross-correlation?
My data are images of shape (512, 1536, 21) so the 2D (x, y) part clearly dominates.
Still, there is a this z-component of length 21. I'm not making use of it so far and I wonder if I should, if it's worth the trouble of finding an implementation and the additional computational cost.
So far, I'm tracking the motion of features on the 2D (512, 1536) images using normalized cross-correlation (function feature.match_template()).
The motion looks mainly planar but, well, I'm not sure how refined I should go about this.