基于高通滤波的频率-空间域经验模态分解压制高频噪声

High frequency noise suppression by Empirical Mode Decomposition in the frequency-spatial domain based on high-pass filter

高频噪声压制是高分辨率地震数据处理中提高信噪比的关键性问题.本文针对f-x(频率-空间)反褶积空间预测滤波器无法处理非平稳、非线性信号的缺点,提出了一种基于高通滤波的频率-空间域经验模态分解(Empirical Mode Decomposition in the frequency-space domain,f-xEMD)压制地震剖面中高频噪声的方法.该方法采用全域高通滤波从原始数据中分离出含有部分有效信号的高频数据,将其变换到f-x域,然后在滑动的短窗口内提取每一个频率的空变数据序列进行EMD分解得到高频复本征模态函数(Intrinsic Mode Function,IMF)IMF1,将所有频率的IMF1序列反Fourier变换到时间域得到噪声剖面,将其与原始数据相减,达到高频噪声压制的目的.该方法可克服传统EMD分解方法中的模态混叠现象,保护陡倾角反射同相轴;压制后的噪声剖面中不包含有效信号能量,地震剖面的信噪比得到了提高.模拟数据和实际数据处理结果充分证明了该方法的有效性.

It is a key issue that high frequency noise is suppressed in high resolution seismic data processing. In order to overcome some shortcomings that f-x （frequency spatial）deconvolution spatial prediction filter can＇t cope with the nonstationary and nonlinear data, this paper presents a method that high frequency noise in the seismic profile is suppressed by Empirical Mode Decomposition in the frequency-spatial domain （f-a^EMD） based on high-pass filter. This method can separate the high frequency data in which include some effective signal from the raw seismic data by the whole region high-pass filter and transform it into frequency-spatial domain For every frequency, the complex high frequency intrinsic mode function（IMF1） is computed from the spatial-varying data in the sliding time window by f-xEMD, then inverse transform all the complex IMF1 back to the time-spatial domain for gaining the noise section. Signal-to-noise enhancement can be achieved by subtracting noise section from raw data. This method can overcome the shortcoming of mode mixing in the conventional EMD method and protect the steeply dipping reflect event. The noise section doesn＇t actually contain the effective signal energy so that this method can restrain the high frequency noise and enhance the signal-to-noise of seismic profile. The processing result of simulated data and actual data indicate that the high frequency noise suppression method by f-xEMD based on high pass filter is very effective for high resolution seismic data processing.