Seismic Wavefield Separation Based on Sparsity Representation

High fidelity wave separation is one of the key issues in process and application of seismic data. A method based on sparsity representation is presented in this paper according to underdetermined blind signal separation theory. It depends on polarization characteristics of seismic wavefield and sparsity of multi-component data in time-frequency domain. The method is a two-step method. Step one, we transform the seismic data from time domain to time-frequency domain and estimate polarization matrix. Step two, we separate pure waves in time-frequency domain and transform them back to time domain. The simulation results show that pure waves are separated accurately and amplitudes are maintained well without flaws such as wave mixing, aliasing and boundary problem, compared with conventional methods.