I want to find a rotated fragment taken from a large image using a fast Fourier transform.
I know how to do this in case the image is not rotated. To do this, it is enough to calculate the similarity function of two images in the frequency domain and find the extremum.
How to solve the problem in the case of a rotated image?