Two-dimensional wavelet-based ground roll filtering

Ground roll is a surface wave energy that propagates along and near the surface with relatively low velocity, often with low frequency, and usually with high amplitudes relative to other events of interest in land seismic surveys. The conventional wave-number (f-k) filtering which is based on the 2-D Fourier transform is not suitable to remove such a time- and space-variant correlated noise. This paper proposes an adaptive 2-D wavelet-based filtering technique by wavelet shrinkage. First, a tree multiresolution decomposition of the data is generated using a 2-D biorthogonal Discrete Wavelet Transform (DWT). Then Level-dependent thresholds are derived based on the Stein's Unbiased Risk Estimator (SURE). Finally, the vertical subbands' thresholds are tuned to remove the ground roll without seriously affecting the refracted events that share the same subbands. The de-noised data is reconstructed by the inverse 2-D DWT.