How can I swap the phases of two fourier transforms?

In my assignment I am supposed to swap phases of the Fourier transform for two images and perform the inverse Fourier transform to reconstruct the images using ifft2(). I am new to Matlab and all I could find was extracting the phase from a fourier transform using angle(), but how do I assign it to another fourier transform?

Answer 1

[Bjorn Gustavsson]

Tip: Think about the difference between Cartesian and polar representation of a complet number are, also be prepared to use element-wise multiplication (i.e.: .*) and the exp-function.

HTH

[Esa Soudbakhsh]

So how does that assign a fourier transform a new phase?

[Bjorn Gustavsson]

You know how to calculate the fft of a matrix? You know how to extract the phase and magnitude of those ffts, now you have to think up a way to combine these 2 arrays from 2 different ffts.

This is your homework, your opportunity to learn, I've now given you the hints twice. If you're brand new to matlab you should have a look at the on-ramp introductions to matlab. Those will get you up to speed. So search for: matlab on-ramp

HTH

[Esa Soudbakhsh]

Alright. From using your hints the only thing that remains is the original question. How do I then replace the extracted phase with the phase of another fourier transform? It's essentially how to use assignment operators in matlab where the only hint would be to google the correct function.

Unless you are saying I should make a new fourier transform from scratch after extracting both the magnitude and phase and placing the new fft in polar form.

[Esa Soudbakhsh]

Okay so what I did was get the phase and magnitude of an fft matrix in rectangular form. Then made a new matrix in polar form by multiplying the new phase with the other magnitude and that made me a new matrix. This is of course after resizing the images to be in the same dimensions.

[Bjorn Gustavsson]

Sounds exactly right! But if you are a bit unsure about the assignment of variables and such I suggest that you browse through the on-ramp courses: matlab-onramp. That is intended to get you up and running as fast as possible.

With regards to the dimensions-nuisance: You might (I've neither thought nor tested this) get somewhere by zero-padding the fft of the smaller images