Attenuating coherent noise by wavelet transform

Abstract

One conventional method for attenuating coherent noise (such as surface multiples and tube waves) is f-k filtering, which assumes there exist pronounced move-out (i.e. apparent velocity) differences between coherent noise and primaries. Considering the non-stationary nature of seismic signals, the potential of wavelet transform as a multi-resolution tool in attenuating seismic coherent noise such as tube waves is investigated in this paper. Using a synthetic crosswell data set, we compare the performance of wavelet filtering and f-k filtering to attenuate strong and aliased tube waves. It will be shown that the wavelet filtering method is superior to f-k filtering in the sense that little filtering noise (such as the Gibbs phenomena) is introduced and primaries are better preserved. Applications to the Glenn Pool crosswell field data sets also show excellent attenuation of the up-going and down-going tube waves and significant recovery of the weak later arrivals.