This paper deals with a discrete time frequency domain approach to super-optimization of four-block problems via a dimension reduction and diagonalization technique based on the so-called equalizer principle, an optimality criterion associated with the polynomial approach to H-infinity-optimization by Kwakernaak. The reduction technique provides a hierarchical decomposition of the super-optimization problem into successive ordinary H-infinity-optimization problems, one for each singular value to be minimized. The procedure allows computation of super-optimal solutions for a large class of four-block problems.