Application of Empirical Mode Decomposition in Random Noise Attenuation
Y. Chen and J. Ma
Event name: 1st Latin American Geosciences Student Conference (LAGSC)
Session: Latin American Geosciences Student Conference
Publication date: 08 April 2013
Info: Extended abstract, PDF ( 829.48Kb )
Price: € 20
Random noise attenuation always plays an important role in seismic data processing. The most widely used denoising method tailored for random noise might be f-x predictive filtering. When the structure of subsurface becomes complex, the filtering method suffers the high predictive error problem becasue of the large amount of dip components to be predicted out. In this paper, we show the limitation of conventional f-x predictive filtering in dealing with multi-dip seismic profiles and present a modified version based on a dip filter enabled by Empirical Mode Decomposition (EMD). We first apply EMD to each frequency slice in the f-x domain and get several Intrisic Mode Functions (IMF), then apply autoregressive (AR) model to the frist IMF to predict the useful steeper events, and finally add the predicted events to the sum of the remaining IMFs. Because of the dip selection property of EMD, the new method can preserve totally the horizontal or low-dip-angle events by filtering out these components and also increase the predictive precision for the high-dip-angle events. Both synthetic and real data sets demostrate the performance of the proposed method.