To date, most of presented approaches of internal multiples prediction or elimination assume an acoustic background model. In practice, the acoustic approximation will result in a false, misleading and potentially injurious estimate due to the neglect of the wave-mode conversion. The closer the reference model is to reality, the more accurate achieved prediction algorithm will be. Inverse scattering series has been revealed in extremely powerful capabilities of seismic data processing and inversion, such as full waveform inversion (FWI), direct non-linear AVO analysis, and surface-related or internal multiples attenuation on acoustic cases, due to its property of model independence. The 3D formulation of elastic internal multiples prediction algorithm was introduced by considering an isotropic-elastic-homogeneous reference medium, which evaluates the wave-mode conversion and multiples prediction in a self-acting manner. The input preparation is the essential step for the prediction. Since the wave-mode conversion can only be handled in the top layer, an inappropriate will misleads the conversion in lower layers which disorders lower-higher-lower relationship of events reflected by the same layer but with different wave mode conversions. In this paper, we analyse the possible input preparation methods for the algorithm implementing in different domains using wavenumber related elastic stolt migration, slowness related elastic stolt migration, time-stretching method in plane wave domain, and the best-fitting method with high resolution hyperbolic radon transform. The advantages and weaknesses of each approach are discussed.
View full article as PDF (0.73 Mb)