Adaptive Empirical Wavelet Transform for Seismic Noise Reduction

Abstract

Summary Denoising is of great importance in seismic data processing. A variety of denoising methods have been proposed in the past several decades. However, it remains a longstanding problem to effectively separate noise and useful signal in the 2D/3D data set. In addition, seismic denoising is often performed trace by trace and the lateral continuity is usually not taken into consideration. To solve these two problems, we propose a novel method named adaptive empirical wavelet transform (AEWT) for seismic random noise reduction. In the AEWT, we take advantage of the outstanding denoising performance of EWT and the lateral continuity of multidimensional seismic data. When dealing with 2D data sets, the AEWT considers the spatial lateral continuity between the traces using the dominant frequency criterion (DFC) method, which selects the intrinsic mode functions by adaptively judging the dominant frequency of each trace. Both 2D synthetic and field data examples show successful performance of the proposed AEWT.