Abstract：The paper uses Curvelet transform in multiple-scale transform system rapidly developed in recent years to replace the elimination process of wave-equation-based predictive subtraction method for free-surface-related multiple suppression, and receives better multiple-eliminated effects. The transform is characterized by optimum sparseness constraint condition that makes the mapping energy of primary reflections on a set of basis function as small as possible. The major implement process as follows: the Curvelet transform is carried out for seismic data, then, to take the coefficients of predictive multiple in Curvelet domain as thresholds by finally using similar noise-eliminated tools to eliminate multiple. The method can meet the minimum energy norm in multiple-eliminated process. In the numeric tests of Curvelet algorithm, the paper tries Radon transform of seismic data first, subtracts the multiple in Radon domain by using Curvelet thresholds method, and multiple-eliminated results are from inverse Radon transform of multiple-eliminated data. The numeric tests showed the better multiple-eliminated effects come from the transform in Radon domain for the Curvelet transform and the energy of primary reflections can be better remained.
