Bernoulli Random Undersampling Schemes for 2D Seismic Data Recovery

Seismic data regularization is an important preprocessing step for seismic signal processing, which has been widely dealt with by compressive sensing recently. Besides sparse representation of seismic signal in some transform domain and 1-norm reconstruction algorithm, the regularization quality depends greatly on random undersampling schemes. For 2D seismic data, discrete uniform based methods have been investigated elaborately. However, designing new undersampling schemes is still an open problem. In this abstract, we propose Bernoulli random undersampling scheme and its jittered version according to Bernoulli distribution law. Experiments with Fourier and curvelet transforms on 2D numerical simulation data illustrate that our novel schemes perform better than or as well as discrete uniform ones.