AUTOREGRESSIVE SPECTRAL ESTIMATION IN ULTRASONIC SCATTERER SIZE IMAGING

An autoregressive (AR) spectral estimation method was considered for t he purpose of estimating scatterer size images. The variance and bias of the resulting estimates were compared with those using classical FF T periodograms for a range of input signal-to-noise ratios and echo-si gnal durations corresponding to various C-scan image slice thicknesses . The AR approach was found to produce images of significantly higher quality for noisy data and when thin slices were required. Several ima ges reconstructed with the two techniques are presented to demonstrate difference in visual quality. Task-specific guidelines for empirical selection of the AR model order are also proposed. (C) 1966 Academic P ress, Inc.