Robust Methods of Adaptive Seismic Multiple Suppression

One of the widely used approaches to multiple suppression is a two-step scheme. In the first step, multiples are predicted. In the second step, multiples are adapted to the input data with matching filters and subsequently subtracted. It is known that a non-stationary adaptation is usually required. A method for computation of smoothly varying matching filters is proposed. The behavior of the non-stationarity is parameterized with decomposition of the variations into a set of smooth basic functions. Besides, substantial improvement is usually achieved by application of multichannel adaptation. Such approach requires a multichannel version of the Levinson’s algorithm, but the computations loose their stability. An iterative robust procedure that does not require regularization is proposed to improve the stability of the scheme.