A high resolution method of seismic data via joint dictionary learning and sparse representation

Abstract

Summary The traditional deconvolution methods have some disadvantages, such as suppressing weak reflection coefficients and are difficult to identify thin interbedding and so on. In order to overcome these shortcomings, this paper presents a new approach to improve the resolution of seismic data, based upon joint dictionary learning and sparse representation (JDLSR). The characteristics of reflection coefficients can be obtained by dictionary learning. In order to explore the correspondence between seismic data and reflection coefficients more efficiently, we introduce the joint dictionary learning. The combined features (DR and DS) of log reflection coefficients and seismic data of well beside can be learned by joint dictionary learning. The known seismic data are sparsely represented under DS to obtain the representation coefficient, which can be combined with DR to reconstruct the unknown reflection coefficients. The effectiveness of the proposed method is verified by the single-channel seismic data and the classical Marmousi model. This method is applied to high-resolution processing of actual seismic data, and it is found that the result is better than sparse-spike deconvolution (SSD).