Introduction

Both surface-related and internal multiples are a major source of coherent noise in many basins throughout the world. Dragoset and Jericevic (1998) describe a two-step multiple elimination process: prediction based on the Delft autoconvolution model Verschuur et al. (1992), followed by subtraction with an adaptive noise cancellation (ANC) algorithm. However, practical multiple removal by adaptive noise cancellation is significantly more complex than the one-dimensional theory described in the text books [e.g. Widrow and Stearns (1985)]. In particular the multidimensional nature of prestack seismic data leads to ambiguity in the choice of potential parameters.

This paper describes an alternative to the conventional ANC algorithm also based on the minimum energy criterion. However, our approach leverages helical preconditioning to ensure the non-stationary shaping filters vary smoothly, and do not accidentally remove primary energy. In our description, we place particular emphasis on determining a robust set of parameters for the process.