反褶积.ppt

谱白化反褶积 Step1:计算地震道自相关的振幅谱 Step2:计算算子的振幅谱 假定算子是最小相位的，则相位谱可以由振幅谱作Hilbert变换计算的得出，即： Step3:计算算子的相位谱 Step4:计算时间域的反褶积算子 Step5:地震道时间域反褶 技术交流 2004/04/20 ② NON-WHITENING DECONVOLUTION 这种类型的反褶积是保留在地震道中的原有子波形态，而不把它压缩成脉冲，本质上它是用来抑制多次波，因此叫预测反褶积，或者叫GAPPED DECONVOLUTION。反褶积算子可以在时间域设计，也可以在频率域实现。 时间域的实现方法 If an operator is designed which does not operate close to time zero we can achieve the objective of suppressing multiples without altering the wavelet. The deconvolution operator is therefore designed to have a Dirac at time zero, a gap containing a series of zero values (hence the name gapped deconvolution), followed by the active part of the filter: calculated by using the WIENER-LEVINSON algorithm. 频率域的实现方法 Let us begin by considering the reciprocity between the frequency and time domains. A smooth amplitude spectrum occurs if the signal in the time domain decays rapidly, that is, it is of short duration. If the amplitude spectrum contains rapid oscillations the signal in the time domain is long: If we now reconsider the amplitude spectrum of the trace autocorrelation function we see that it does indeed contain oscillations: The undulations in the spectrum can be considered to be due to the presence of reverberations. A smoothed curve drawn through this spectrum can therefore be assumed to be the amplitude spectrum of the source wavelet: In non whitening deconvolution the operator is designed to remove the reverberation components from the spectrum but to leave the smooth curve difference between the smooth and unsmoothed spectra operator therefore： 技术交流 2004/04/20 ③ MULTI-CHANNEL DECONVOLUTION 常规反褶积是逐道设计的算子（地表一致性反褶积除外），在这种情况下有两个缺陷： 噪音问题 如果地震记录上的噪音道简变化很大，用常规逐道方法设计的反褶积算子在相邻道间也会有很大变化。这将导致降低地震记录上同相轴连续性的质量。 信号问题 在有些地质情况下，一次反射周期之间不具有随机性，或者说某个一次反射周期与多次波时间一致，这样一次反射会随多次波一同去除。 A solution to the problems caused to conventional deconvolution by changing noise patterns and non-random geol