The Radon transform is defined as summation along some specified family of trajectories over a set of time domain seismic data. The hyperbola is the most interesting summation curve for NMO-corrected seismic CMP-gathers since primaries become flat, but multiples remain hyperbolic. The primaries and multiples are hence able to be separated in the hyperbolic Radon domain. However the Radon transform is not a one-to-one transform, which means that near offset energy is transformed to the Radon domain many times. When we apply the inverse transform, the amplitudes of the reconstructed data are actually different from those of the original input data. In this paper, the strongest events on a record are first estimated by the semblance weighted Radon transform, and the corresponding energy is removed from the input data, forcing a forward Radon transform to be one-to-one. Removal is actually accomplished by a nonlinear filtering of multiples from the original input data rather than by subtraction.
View full article as PDF (0.22 Mb)