Restoration of scatterer distribution based on Empirical Bayesian learning with consideration of statistical properties

In ultrasound medical imaging, speckle patterns caused by reflection interference from scatterers in living tissue are often suppressed by various methodologies. Accurate imaging of scatterers is important in diagnosis, since a structure of scatterers distribution has information on tissue properties. Therefore, it is necessary to solve the band limit imposed on the echo mainly by the characteristics of the transducer. Since scatterers are spatially correlated and thereby constitute a microstructure, we assume that scatterers are distributed according to the autoregressive (AR) model with unknown parameters. Under this assumption, scatterers distribution is restored based on the empirical Bayesian learning. That is, the AR parameters are estimated by maximizing the marginal likelihood function, and the scatterers are estimated as a maximum a posteriori (MAP) estimator using these. Such a scheme is stably realized by the expectation-maximization (EM) algorithm. The method based on the empirical Bayesian te...