【摘要】 预测误差滤波PEF(PredictionErrorFilters)模型有一个隐含的假定 ,即误差是源噪声。f -x预测滤波是用一个震源噪声来计算一个误差剖面 ,并把这个误差剖面作为一种可加噪声从记录中减掉来实现去噪的目的。因此 ,这与误差是源噪声的隐含假定是不一致的。研究讨论一种 f -x投影滤波方法可以避免产生这种不一致性。在非线性方程组迭代求解中给出了一种稳定化的初始值选择方法 ,在求取最小相位信号时给出了一种矩阵分解方法。通过理论模型和实际地震记录的计算 ,表明了这种去噪方法的有效性更多还原

【Abstract】 The PEF(prediction error filters) model implicates that the error is a source noise. f-x prediction filtering computes an error section as a source noise and removes it from the record section to obtain the signal section. In the second process above, the error section is obviously used as an additive noise. It is therefore model-inconsistent, f-x projection filtering(PJF) can avoid the model-inconsistency. In this paper, the PJF method is discussed. A stable method on choosing initial values is given for solving a nonlinear system of equations with an iterative process. A matrix decomposition technique is employed to compute a minimum phase signal. Computing results, both for a synthetic seismogram and for a real seismic record, show that the method is effective.更多还原