2-D state-space filter structures with low frequency-weighting sensitivity

Based on the Roesser local state-space model, the coefficient sensitivities of two-dimensional (2-D) digital filters are analyzed over a particular frequency region. The overall sensitivity, called the frequency-weighting sensitivity, is then evaluated using the 2-D generalized Gramians. Next, 2-D filter structures that minimize the frequency-weighting sensitivity are synthesized for two cases of no constraint and scaling constraint on the state variables.<
>