Ground roll attenuation based on a hybrid wavelet transform and singular value decomposition
Rosita Hamidi, Abdolrahim Javaherian, Ali M. Reza · 2012
Ground roll is of great importance in seismic records acquired in reflection surveys on land because of its unique properties. Ground roll with linear coherency and stronger amplitude than reflected signals can totally cover reflections and mask the valuable information carried by the signals. The main concern in applying filters based on the wavelet transform (WT) is to differentiate between the coefficients corresponding to noise and signal. In the case of ground roll attenuation, since it is a low-frequency coherent noise, coefficients at higher scales and specific time spans - depending on trace offset - are considered as the noise coefficients. Ground roll also has higher amplitude than signals (reflections) which results in greater wavelet coefficients. In this paper, singular value decomposition (SVD) is used as a criterion to select coefficients representing the noise. The results of the ground roll attenuation in the synthetic and real data show that parameters controlling the filter are distinct frequency band, arrival time, and energies of noise and reflections.