Symmetrical decomposition and transformation (Q1064307)

Some procedures to design two-dimensional (2-D) digital filters are presented. The filters under consideration are of finite impulse response (FIR) type. As the starting point two main facts are considered: the first is the so-called McClellan transformation (MT) which allows to obtain a 2-D digital filter starting from 1-dimensional filter; the second is the way in which some symmetry properties are carried on from the transfer function of the 1-D digital filter to the transfer function of the new 2-D digital filter when using the MT. Next, coefficient conditions stemming from various symmetries of real two-variable functions are derived, and it results an important decreasing of the number of the distinct coefficients. An important result is contained in Theorem 8 which gives necessary and sufficient conditions for a real two- variable function to be decomposed as the sum of the symmetrical and antisymmetrical part of some operation T. The main problem which is finally asked is the following: if, instead of optimizing the coefficients of a 2-D filter, obtained by the MT, one first decomposes the transfer function into some parts exhibiting particular symmetries and then optimizes the coefficients of the components, is this way leading to less computer time? Some examples seem to give an affirmative answer but the general answer is still to be found. The paper is very clearly written, the main ideals are illustrated by interesting examples, and this is a fair mathematical investigation in the area of digital filter design.