Bit-serial generalized median filters

Generalized median filtering techniques that have appeared in literature suffer from some severe disadvantages. They are not only hardware intensive and time consuming but also tend to smear image edges. These shortcomings can be overcome by having bit-serial filtering techniques. The median selection of the sample values in the window is done for a few most significant bit stages. The sample values not in consensus with the majority are edited based on some criterion and averaged out with the rest of the data to yield the filter output. Depending on the criterion for editing the data, two filtering techniques-the B filter and the Tag-based Trimmed Mean (TTM) filter have been presented. The median-running mean filters are integrated in an area efficient manner to provide the basic building block for a signal processing chipset