Multi-dimensional power spectrum estimation using noncausal rational models

Methods are presented for the identification of noncausal, rational, multidimensional systems from covariance data in connection with the development of noncausal models in multidimensional power spectrum estimation. It is shown how a recently proposed notion of state for noncausal systems and the resulting rank properties can be used for model estimation. The general class of noncausal systems studied encompasses the quarter-plane causal, all-pole, separable, and factorizable models previously considered for 2-D spectrum estimation