Efficient algorithms for three-dimensional filtering and linear prediction

In this paper optimal MSE three-dimensional (3-D) filtering and linear prediction is considered. 3-D filter masks of general shape are allowed. An efficient multichannel embedding is utilized that transforms the 3-D model to an equivalent multichannel formulation. Efficient recursions are developed for updating of lower order filter parameters towards neighboring points. Fast algorithms are applied for the solution of the 3-D normal equations in an order recursive way.