An algorithm for designing constrained least-squares filters

This paper describes a practical computer algorithm for the solution of constrained least-squares filtering equations. This algorithm exploits the block-Toeplitz and block-circulant properties of the filtering equations. Specifically, these properties are utilized to adapt Kutikov´s and Akakai´s algorithms for the solution of block-Toeplitz systems. Our approach generalizes the formulation considered by Claerbout. Furthermore, it effects substantial savings in memory storage compared to existing approaches, and is therefore well suited for solving large systems.