Frequency Estimation of 2-D Sinusoids from Very Limited Data

A novel algorithm for estimating the frequencies of nearly maximum number (as bounded by the number of equations vs. unknowns) of 2-D sinusoids is described. The method is parametric model based and therefore it is not Fourier resolution limited. It utilizes redundant data in a least squares manner, therefore is quite robust and approaches asymptotic efficiency relatively fast. Finally, it is computationally affordable and does not require 2-D polynomial rooting. A numerical example of retrieving 12 2-D sinusoids from a 6times5 grid is demonstrated