Abstract
In this paper we propose a new two-dimensional (2-D) spectral estimation method which has high-resolution property even in the case where the measured 2-D data is of short length and at low signal-to-noise ratio. The method combines the iterative filtering method, addressed by the authors, with a data extrapolation technique. The power spectrum is estimated by 2-D periodogram after the data extension based on Burg's AR spectral analysis.
The iterative prefiltering reduces noise components embedded in the data, making the following data extrapolation accurate. As a result, we can obtain high-resolution spectral estimates from the extended data.
Computational complexty of the method is moderate, because the iterative prefiltering is basically calculated by 2-D FFT algorithm, and the data extrapolation is done based on 1-D Burg's order-update algorithm.
Computer simulations show the effectiveness of the method.
