Nonlinear fusion filters based on prediction and smoothing

To attenuate white noise, nonstationary noise and impulse noise are importa nt for signal processing. In this letter, we present nonlinear fusion filte rs (NFF) based on prediction and smoothing. By means of least square fittin g of a polynomial, we define and give the operators of left prediction and right prediction, left smoothing and right smoothing, central smoothing and cross-validation smoothing. in simulated experiments, it is shown that the present method is an effective one.