Removal of Random Noise in Seismic Data by Time-varying Window-length Time-frequency Peak Filtering

Abstract

Time-frequency peak filtering (TFPF) is an effective tool for the removal of random noise and can be used to process seismic data with a low signal- to-noise ratio. A crucial aspect of this algorithm is the choice of window length (WL) of the time-frequency distribution. Whereas a fixed WL cannot simultaneously preserve signal and attenuate noise, timevarying WLs can achieve this goal. We propose a new method, L-DVV (delay vector variance), which successfully processes non-stationary signals by using the surrogate to measure the non-linearity of a time series. This method is sensitive to random noise and can accurately recover seismic signal masked by noise. Since the linearity criterion also meets the unbiased estimation criterion of the TFPF algorithm, the L-DVV method can be used for time-varying WL TFPF processing. Analysis of synthetic and real seismic data shows that the time-varying WL TFPF algorithm is effective at removing noise and recovering seismic signal.