Simplifying Standard Median Filtering Algorithm by Window Decomposition and Medians Combination Filtering

摘要

As the properties of edge preservation and efficient noise attenuation, median filter is widely used. For data processing in low-cost hardware platform such as wireless sensor network nodes, the filtering algorithm must be simple and efficient. In order to reduce the calculation complexity of the median filter with larger window size, a new filtering algorithm which trades off between calculation complexity and noise attenuating ability is proposed. Based on the hierarchical structure of median hybrid filters and applying standard median filtering, the algorithm divides the filtering window into three sub-windows and properly selects the samples for each sub-window, then combines the medians of the subwindows and outputs the median of the combination. The simulation and calculation complexity analysis show that the simplified filtering algorithm effectively reduces the computations, retains the noise attenuating and edge preserving ability and is applicable to low-cost hardware.