一种具有频率选择特性的加权伪中值滤波算法

相对于中值滤波而言,伪中值滤波具有计算快速等特点.通过引入负加权系数并使滤波窗口大小无奇偶限制,该文将原有的伪中值滤波扩展为具有负系数的加权伪中值滤波.基于Mallows样本选择概率理论,由FIR滤波器的脉冲响应来获得伪中值滤波器加权系数,使其具有和FIR滤波器一样的频率选择特性.仿真结果表明,本文提出的滤波方法在赋予了伪中值滤波频率选择特性的基础上,减小了计算量,保持了中值滤波所具有的优点,并且能够去除信号中高频的周期性干扰噪声.

A New Frequency Selective Weighted Pseudo-median Filter Algorithm

The pseudo-median filter is a computationally efficient alternative to the median filter. This paper extends the definition of pseudo-median filter when the size of filter windows is even, and proposes the Negative Weighted Pseudo-Median Filter (NWPMF). The NWPM uses the Mallows’ theory of sample selection probability to design the weight coefficients according to the respective impulse response of FIR filter with the desired spectral response. Because the weight coefficients can be negative or positive, the NWPM is characteristic of the frequency selection. The simulation results show that the NWPM can reduce the computations and is superior on reserving fine details, removing impulsive noise and high frequency periodic noise.