Adaptive preconditioning of full waveform inversion based on structure-oriented smoothing filter

초록

In seismic full waveform inversion, regularization or preconditioning have been exploited to stabilize the inverse problem. Isotropic smoothing is one of the easiest way to implement regularization in full waveform inversion. However, the isotropic smoothing has a drawback that it blurs layer interfaces such as faults, sedimentary layers and salt boundaries. Consequently, the application of a structure-oriented regularization based on geological information is required to recover accurate subsurface structures. In this paper, we propose an adaptive structure-oriented preconditioning using diffusion tensor and weighted semblance. Our preconditioning algorithm is to apply anisotropic smoothing based on structural information obtained from migration image. By using weighted semblance, structure-oriented smoothing is much more effectively conducted. We apply the proposed structure-oriented preconditioning to the full waveform inversion of the Marmousi and SEG/EAGE salt models. The structure-oriented smoothing filter is periodically and adaptively updated as the inversion progresses. The inversion result shows that the layer interfaces are faithfully preserved and sharpened. © 2018 SEG.