Model reduction of 2-D separable-denominator transfer functions via quasi-Kalman decomposition

In this paper a model reduction algorithm for casual recursive separable denominator (CRSD) 2-D digital systems is proposed. The algorithm is based on the quasi-Kalman decomposition (QKD) method developed for 1-D systems. The proposed algorithm employs singular value decomposition of the horizontal and vertical Hankel matrices. This leads to producing two sets of finite Gramians where in each set the finite Gramians are equal and diagonal. An example is given in order to illustrate the method.