基于最大似然准则的f-x-y域l_1模预测滤波方法

Prediction filtering method using l_1 norm based on maximum likelihood rule in f-x-y domain

常规的f-x-y域预测滤波方法基本上都假设随机噪音是服从高斯分布的,但在实际地震资料中并不是所有的随机噪音都服从于高斯分布。针对非高斯分布的随机噪音,讨论了f-x-y域l1模预测滤波方法。该方法利用最大似然准则建立目标函数,同时利用非线性共轭梯度法来优化求解目标函数。理论模型试算表明,f-x-y域l1模预测滤波方法不仅可以有效地消除非高斯分布的随机噪音,而且也避免了假同相轴现象的产生。

It is supposed that the random noise obey the Gauss distribution in conventional filtering methods, But some kinds of random noises in the real seismic data do not obey the Gauss distribution. In order to eliminate the random noise which obeys the non-Gauss distribution, the prediction filtering method using l1 norm was discussed in f-x-y domain. The objective function was established based on the maximum likelihood rule, which was solved by nonlinear conjugate gradient method. The results of theoretical model test show that the prediction filtering method using l1 norm can not only effectively eliminate the random noise which obeys the non-Gauss distribution, but also avoid the artificial event.