Rank identification for an analog ranked order filter

The paper suggests a method for rank identification from a previously reported programmable parallel current-mode analog ranked order filter circuit. This allows the input direction with any selected rank to be indicated as well as extracting the correct value from a set of input currents. The rank identification can be implemented with only O(n) increase in circuit complexity. This addition can be used to avoid errors caused by device mismatch and may also result in new application methods for the ranked order filter. Programmable ranked order filtering can be used for many image processing tasks such as mathematical morphology and median filtering.

I. Introduction

Ranked order filtering, which means extracting from a set of inputs, the one with the selected rank, is the generalized case of order statistic filtering. Usually the most important special cases of order statistic filtering are the extraction of the minimum, maximum or the median from a set of input values. These operations can be used for a variety of grayscale image processing functions such as median filtering, weighted median filtering and image processing techniques based on mathematical morphology, which require local minimum and maximum operations. In the case of soft morphology, also more general order statistics may be needed [1].