On root structures and convergence properties of weighted median filters (Q1908848)

Some root properties of the weighted median filter are studied. The necessary and sufficient conditions for any one-dimensional signal to be a root signal of these filters are presented. Some examples are given to show these conditions explicitly. The convergence of the symmetric weighted median filters is studied. The analysis is restricted to periodically time-varying inputs. The obtained results are proven for a subclass of weighted median filters, the class-1 filters. After defining an important parameter, the feature value, the authors show that any one-dimensional unappended signal of length \(-L\) will converge to a root signal in at most \(3[(L-2)/ 2(2N+ 2-p)]\) passes through a class-1 filter with window width \(-2N+1\) and feature value \(-p\). The paper will be of interest for the specialists and engineers, working on LMS digital filter design and their VLSI implementations.