A split sorting-algorithm for generalized order statistic filters

A split sorting-algorithm which possesses excellent parallel and pipelined characteristics and can be used to implement a class of generalized order statistic (GOS) filters is introduced. The algorithm is first developed for the computation of order statistics (OS), and its parallel/pipelined properties are discussed. The algorithm is extended to accomplish the task of sorting a whole data group in a parallel manner. The computation of running OS using the developed algorithm is demonstrated to be simple and efficient. As compared to other fast sorting-algorithms, it is noted that certain advantages can be obtained by using this algorithm. It is shown that the ultimately split structure will lead to a parallel and pipelined realization which requires less than linear complexity