Seismic erratic noise suppression based on Laplacian-scaled mixture prior

Noise suppression is always a vital procedure in seismic data processing. Under the Gaussian distribution assumption, conventional sparsity-based methods are highly practical and effectual in dealing with random noise. Nevertheless, the field data are always contaminated by erratic noise with high amplitude and unknown distribution. Here, to handle the erratic noise, we present a Laplacian-scaled mixture based robust sparse representation method. The Laplacian-scaled mixture (LSM) is used to model the erratic noise under the Maximum A Posterior (MAP) framework. The random noise is characterized by Gaussian prior. In addition, the sparsity of the clean data is given by the Laplace prior. Via alternative iterations, the parameters can be solved effectively. The effectiveness of our method is presented using synthetic and field examples.