2-D recursive bilinear model for nonlinear signal representation

Develops 2-D recursive bilinear models for quarter-plane and weakly causal random fields. These models are obtained as a nontrivial generalization to two dimensions of the 1-D bilinear time series and satisfy a 2-D bilinear recursion where the coefficients lie on a causality cone. The extended support of the coefficients results in greater model flexibility. An efficient maximum-likelihood estimation method for order determination and parameter estimation has been obtained. Computer simulations are included to illustrate this model and the performance of the parameter estimation technique.<
>