High-fidelity Adaptive Curvelet Domain Primary-Multiple Separation

In this paper, we propose an adaptive implementation scheme for first separating multiples from primary events by a given multiple model in seismic data and subsequently removing the multiples from noisy seismic data using the curvelet transform. Due to the sparseness of seismic data in the curvelet domain, the optimization problem is formularized by incorporating L1- and L2-norms, based on the framework of Bayesian Probability Maximization. Moreover, to meet the challenges faced by various types of data complications, we further develop a method termed frequency-regularized adaptive curvelet domain separation for enhancing the effectiveness of primary-multiple separation by performing frequency dependent optimization in response to the presence of noise and the inaccuracy of multiple models.