The influence of limited linearization of time window on TFPT under the strong noise background
强随机噪声条件下时窗类型局部线性化对TFPF技术的影响

Time-frequency peak filtering (TFPF) method is a staple technique for eliminating strong random noise in seismic exploration data. For applying these techniques extensively, this paper discussed the details of application of this method, such as choice of the length of time window, local linearization in one time window, etc. For the results of the simulation of time-varying time window TFPF, we compare many values, including the strength of global background, the amplitude of spectrum, signal-to-noise ratio, mean square error, the amplitude of peak/valley of valid wavelet, and the degree of the aberrance of wavelets, and then conclude synthetically that time-windowing filtering (L0+Lx) will get rather ideal effect. In the case of triangle waves, the filtering effect of TFPF is related to the average curvature of side line, the length of time window, etc. The bigger the average curvature of side line is, the bigger the mean square error is; moreover, increasing the average curvature of side line will cause a trend of decreasing in the chosen length of time window. In conclusion, applying TFPF method for eliminating random noise in seismic exploration, suitable parameter of time window should be chosen, on the other hand, the parameter of time window should not vary too much, so the waveform of valid wavelet is not distorted.