Estimation of the frequency vector of a two-dimensional harmonic signal using three-point interpolation algorithms and singular value decomposition

Abstract

Computationally simple algorithms are proposed to estimate the local frequency vector of a real two-dimensional signal with slowly varying parameters in the presence of additive Gaussian noise. The rough most likely estimate of the frequency vector is refined using three-point interpolation algorithms and the singular value decomposition of a 3 × 3 matrix of discrete Fourier transform coefficients. The results of computer simulation are presented showing the efficiency of the algorithms for analyzed fragments of small size.