Sorting networks using nonlinear L/sub p/ mean comparators
Michail Angelos Pappas, Ioannis Pitas · 2002
In certain signal processing applications there is a need for fast hardware implementations of sorting algorithms and networks. So far, classical minimum/maximum comparators have been utilized in various sorting network topologies. However, these comparators can not attain high speeds in operation, due to limitations in digital technology. This paper introduces the L/sub p/ comparators, which are based on the theory of nonlinear mean filters. It is shown that the disadvantage of introducing errors is counter-balanced by their faster performance, when compared to the performance of classical comparators. A novel L/sub p/ comparator-based sorting network is also presented, for the fast calculation of the median of a data set. In this implementation, the number of steps required to produce the ordered output is not related to the number of inputs.