A median filtering method with directional and time-space variant haracteristics
Sun Zhe*, Cao Wuxiang, Jianfeng Wang, Peng De-li · 2014
As a simple and effective nonlinear smoothing filtering technique, one dimensional classical median filtering method has been widely used in seismic processing. In the process of noise suppression, one dimensional classical median filtering method is normally applied along only one direction. However, the dip of the events in the seismic records are always varying. This will degrade the effect of noise suppression. Besides, the filtering length is fixed in this method, but in the seismic records, the SNR is usually varied with time and space. Therefore, when processing the seismic record with this method in the location with strong signal, the signal may be over damaged, and in the location with strong noise, the noise may be suppressed insufficiently. Aiming at the above problems, on one hand, a method of selecting the appropriate filtering direction by the coherence characteristic of the event is used; on the other hand, a method of setting the appropriate filtering length which is time-space variant according to the degree of signal coherence is employed. Based on two above aspects, an improved median filtering method is proposed in this paper. The processing results of both the theoretical model and the real data show that this improved median filtering method in this paper can protect the signal well while eliminating the random noise effectively.