Imagine you have a real image, and you put it through a 2D FFT.
Usually this yields a cross-like structure (edge effect) and some real content, depending on the image.
Imagine the original image contains two spots with bad lighting, which for example yield two different frequencies in the fourier domain.
Imagine another, different image, put through fft, with a single spot due to bad lighting; this spot yields the same frequencies in the fourier domain, as the spots in the first image combined.
How would one distinguish those two amplitude spectra? In my opinion, there is no way of knowing the location of a certain frequency in the fourier domain. Only a direction information is retained, i.e. a horizontal line in the image will yield a vertical line in the fourier amplitude spectrum.
So the information I am after has to be hidden in the phase spectrum. How can I recover such a specific piece of information from a phase spectum that looks like noise?
