SPARSE DICTIONARY LEARNING FOR SEISMIC NOISE ATTENUATION USING A FAST ORTHOGONAL MATCHING PURSUIT ALGORITHM

摘要

Attenuation of random noise is a long-standing problem in seismic data processing. One of the most widely used approaches is based on sparse transforms. In the geophysics community, most of the currently used sparse transforms have fixed bases, which we call analytical transforms. In this paper, we seek a different type of sparse transform, with variant bases, to attenuate random noise. We call this type of transform dictionary learning-based (DLB) sparse transforms, because it can adaptively train a sparse dictionary from the observed data to adapt to different seismic data. To increase the efficiency of sparse dictionary learning, we propose to apply a fast orthogonal matching pursuit (OMP) algorithm for sparse coding. We use both synthetic and field data examples to show the superior performance of the dictionary learning-based transform over fixed-basis transforms, and much improved efficiency in sparse coding associated with the fast OMP algorithm, which is one of the two steps in the DLB transform.