A VLSI architecture for the implementation of real-time order statistic filters

The authors give an algorithm for the real-time computation of the running order statistic (ROS). Order statistic filters (OSFs) or ranked order filters (ROFs) are a class of nonlinear digital filters which have found applications in image processing and in smoothing of speech data. Real-time implementation of an OSF requires the computation of the order statistic (ranked order) of the samples in a window (usually containing an odd number of samples) which gets periodically updated with the arrival of a new sample(s). A modular and highly parallel architecture characterized by programmable window size and required rank order is proposed for the implementation of the algorithm in VLSI form