A modification of a scheme developed previously is presented. The technique uses the McClellan transformation applied to a 1-D zero-phase recursive filter to obtain a 2-D zero-phase recursive filter which is unstable. The stabilization process is done through the decomposition of the 2-D zero-phase filter using the spectral factorization technique into four single quadrant filters. The coefficients of the derived single quandrant filter can be used as the initial values in a direct optimization scheme to minimize an error criterion, which includes the phase and magnitude response of the filter.