Deconvolucao De Reflexoes Multiplas E Peg-Legs Utilizando Filtragem Wiener-Levinson Multicanal

In the present paper we have used multichannel predictive<br>filters of Wiener-Levinson (WL) type to deconvolve<br>surface and internal multiples and peg-legs. The<br>WL multichannel filters are obtained as a solution for<br>block-Toeplitz systems of normal equations. These systems<br>are formed with coefficients of auto and crosscorrelations<br>of a set of seismic traces. The obtained<br>filters can explore in a more efective way the in-time<br>and spatial redundancy of informations that exists in<br>the seismograms. In order to place the seismic data<br>in a more appropriated way to be deconvolved we have<br>applied the MMO transform. By forming panels of<br>common-offset the WL multichannel filters were applied<br>in cascade. Numerical examples using marine seismic<br>data from the Gulf of Mexico illustrate the performance<br>of the method. The results have been quite satisfactory,<br>thus demonstrating that the method is robust and effective<br>to remove the multiple and peg-leg reflections.