## Super resolution using frequency domain.

Started by 6 months ago151 views

Hi,

I am trying to code in MATLAB a super resolution algorithm that was published in 1984. ("Multiframe Image Restoration and Registration" by Tsai and Huang 1984). I have attached the main part of the paper to this post:

01.png

This algorithm assumes a known translation in the x y plane between each “low resolution” frames. It then uses the Fourier transform and some its properties to rebuild a single high-resolution image.

From what I understand, G(m,n) = phi(m,n)*F(m,n) [equation 10] needs to be solved for F(m,n) for each frequency pair m,n. The ideal image can then be reconstructed by calculating the inverse FT of F. G(m,n) should contain all the FT coefficients of all the low resolution frames at the particular (m,n) set of frequency. I am specifically struggling with the matrix phi, I am not sure how to calculate it.

My questions are:

• How do I calculate “s” and “gamma” used in equation 11 and what do they correspond to?
• What should I do with “F” after solving equation 10?