0

I want to estimate an affine transform given matches between a transformed img and a reference img (which came from a reference stack).

Can I estimate a 3D affine transform given only the x,y coordinates of the transformed image and the x, y, z coordinates of the reference image (z being the slice from the reference stack that the reference img came from)?

The general formula for illustrating a transform is:

x' = M * x, where x' is the transformed point. M is the transformation matrix, and x is the original point. The transform matrix, M, is estimated by multiplying x' by inv(x).

The standard setup for estimating the 3D transformation matrix is this: enter image description here

How can I estimate the transformation matrix if I don't have the z' of the transformed image? Is there some other setup I can use?

haxtar
  • 1,962
  • 3
  • 20
  • 44
  • Does it really make sense to include the z-coordinate? After all, it will be equal for all pixels. Or do you want one transform for all images of the stack? – Nico Schertler Aug 17 '16 at 21:30
  • Do you want to find the 3Daffine transformation between a slicing of a 3D volume(meaning that the transform img goes through several different slices in the stack) if yes i thinkk you should give the points you find in the transform image z=1 but you find the best corresponds from all the points you find in the entire stack and not only from one slice. – Amitay Nachmani Aug 18 '16 at 14:07

0 Answers0