The side-information median filter for sequence reconstruction

The side-information median filter (SIMF) is introduced as an extension of the standard median filter (MF) for sequence reconstruction purposes. The SIMF uses the filter input sequence index information for determining the filter output sequence index of an equivalent MF output. The set of SIMF roots is shown to be larger than that of MF roots, including many bounded variations which would be distorted by the MF. Several deterministic and statistical properties are derived for the SIMF, and the SIMF is shown to preserve details in many more sequences than the MF while still maintaining noise impulse-rejection properties similar to those of the MF. Experimental results are presented from sequence reconstruction tasks involving the SIMF, the MF, and other sequence reconstruction filters. Results for both 1-D and 2-D sequences are reported.<
>