基于时空变中值滤波的随机噪声压制方法

Random noise elimination based on the time-space variant median filtering

针对一维中值滤波的滤波方向与滤波窗长单一的缺陷,一方面根据反射信号沿同相轴方向的相干特征,分别对每个采样点设置合适的滤波方向;另一方面根据地震数据的相干程度,分别对每个采样点设置合适的滤波窗长,从而得到具有时空变特性的改进中值滤波方法。理论模型与实际数据的处理效果均表明,无论是对于理论模型合成的单炮记录还是波场复杂的实际地震资料,本文方法均能在有效地压制随机噪声的同时,较好地保护有效信号。

Classical one-dimensional median filtering methods have some defects such as singleness in filtering direction and filter window length.To solve these problems,we propose in this paper a method for random noise elimination based on the time-space variant median filtering.On one hand,we set an appropriate filtering direction at every sample point according to the coherence characteristics of reflection events;and on the other hand,we choose an appropriate filtering length at every sample point according to the coherence degree of seismic data.Test results on model and real data show that the proposed method can effectively eliminate random noise on both shot records and seismic sections.