Adaptive ground roll attenuation in curvelet transform domain
Bahareh Boustani, Siavash Torabi, Abdorahim Javaherian · 2012
In seismic records, the desired objects are disturbed by noise. More separation of the noise and the signals results in better noise attenuation. This may not be accomplished well in a time domain and classic transform domain. A new multiresolution and multidirection domain which is proper for this purpose is the curvelet transform. This transform divides the 2D f-k domain into various scales, directions and positions. These characteristics cause a good separation of the ground roll and the reflectors in the curvelet domain. Using a non-linear approximation, the curvelet coefficients in specific scales and angles can be set to zero. Non-linear approximation can be applied by thresholding. In this paper, using the energy thresholding, the ground roll was removed from the curveletsubbands. In the synthetic data after applying the adaptive curvelet, most of the low-frequency ground roll attenuates. Nonadaptive curvelet attenuate the ground roll better than the adaptive ones. However, it is not necessary to observe all angles in each scale of the curvelet domain for detection of the ground roll subband.