Data extrapolation for high resolution radar imaging

In radar imaging, AR modeling is sometimes used to extrapolate the scattered field data to obtain a high resolution image. In general, the Burg method is used to estimate the prediction parameters. The Burg method leads to a stable prediction filter but can also cause bias in the estimated spectra. One can also use the modified covariance method (MCM) to estimate the prediction parameters. These parameters lead to unbiased spectra. However, the MCM does not guarantee a stable prediction filter. One may have to modify the prediction parameters to ensure a stable prediction filter. One way to ensure stability is to reflect the unstable poles inside the unit circle. It is shown that the modified parameters can be used effectively for data extrapolation. The radar images obtained using this extrapolated data are more accurate than those obtained using the extrapolated data from the Burg prediction parameters. >