Technical info on transformation

Hi-- thanks for all your work on this project. I can not find information on how your algorithm transforms (warps) the new image to align with the old one. Is it scaling and rotating, like a Helmert transformation? Or are the horizontal and vertical scalings separate, like a polynomial transformation? Or does it depend on how many point pairs I create?


1 Like

Hello Robert,

we compute an optimal homography, i.e. a perspective transformation between two planes, with four or more points. With four points an unique solutions exists, for more points an optimal solution

needs to be computed. For a slightly different approach you may refer to our paper:



1 Like