Curvelet noise attenuation with adaptive adjustment for spatio-temporally varying noise
Bogdan Kustowski, Jeffrey Cole, Harry Martin, Gilles Hennenfent · 2013
Curvelet noise attenuation (CNA) has proven to be an excellent technique for suppression of incoherent, as well as coherent noise in seismic data. The basic implementation of CNA involves thresholding of the coefficients in the curvelet domain and it can handle only data with a relatively constant level of incoherent noise. Since this requirement is rarely satisfied by seismic data, trace amplitudes are often normalized using an Automatic Gain Control (AGC) prior to CNA. We present an alternative approach, which simplifies and accelerates the workflow and eliminates the need to store additional copies of data. The new algorithm estimates spatio-temporal noise variations in the curvelet domain and uses this estimate to modulate the coefficient threshold. We demonstrate that the new algorithm works well on data with gradually varying amount of incoherent noise and may lead to smaller signal bias compared to the CNA with prior AGC normalization.