Self-tuning two dimensional predictors: transfer function approach

Summary form only given. Results from self-tuning control have been translated and extended to the area of two dimensional prediction. Specifically, a transfer function approach is used whereby the Diophantine equation used to construct the optimal k-step-ahead predictor in one dimension is extended to two dimensions. In two dimensions, the solution is a rational transfer function and hence the predictor algorithm can be implemented in closed form. This has the further advantage that the variance of the prediction error may also be computed in closed form. The predictor design work is placed in a self-tuning format by combining the two-dimensional ARMA representation of the whole field