Combining Pseudo-median Filter and Median Filter to Improve Performance

摘要

Despite many of the same properties betweenstandard median filter (SM) and Pseudo-median filter (PM),SM can remove impulses completely but PM not. However,PM can be more efficiently calculated which is very importantfor processing data online in WSN. In order to synthesize theadvantages of performance of PM and SM, an improvedmethod combined the advances of PM and SM is proposed inthis paper. A soft-switching scheme based on noise detection isdeveloped to select the appropriate filter for each processing.In the noise detection each sample is classified to containedimpulse or normal by increment contrasts of latest two outputsand between the current sample and the latest. Then SM andPM are alternatively used to process the impulsive samples andthe normal. The simulation results and calculation complexityanalysis show the combined algorithm attenuates impulsesmore efficiently than PM and can be more easily calculatedthan SM. So it can be seen as an alternative of high efficiencyfor applications in wireless sensor network.