New approach for an approximate realization of rank order filtering with low complexity

In radar systems rank order filtering is used in order to distinguish interesting signals from noise, clutter and interferences. For implementation various algorithms based on sorting procedures or histograms are known typically requiring high complexity in computational load and/or data storage. This paper introduces a new approach for an approximate realization of rank order filters with low complexity. Basis idea is subdividing the total range of possible values in intervals suited to the used number representation, establishing a histogram over these intervals with the quantity of values therein, accumulating the quantities, determining the interval with the searched rank, and approximating the searched value by interpolation in this interval. The quality of approximation can be chosen as good as desired.