Full Waveform Inversion based on Gradient Bidimensional Empirical Mode Decomposition

Summary Full waveform inversion (FWI) is a promising technique for high-resolution velocity building. FWI contains both the tomography and migration component. For reflection data, the tomography component update is much weaker than the migration component. When the FWI gradient is dominated by the migration component, the inversion is easy to fall into local minimum, which leads to bad result at deep part. So, it's necessary to separate the tomography component from the original FWI gradient to build a good background velocity. In this abstract, we propose to use the Bidimensional Empirical Mode Decomposition (BEMD) to separate the original FWI gradient into the tomography and migration component. The BEMD can adaptively decompose the gradient into different scales components. The small-scale component can be combined into the migration component and the large-scale component can be combined into the tomography component. In the beginning stage of FWI, the background velocity is updated with tomography components, and with iteration, small-scale components are gradually added to update the velocity model. Numerical experiments show that the proposed method can reconstruct the background velocity model effectively, and improve the accuracy of inversion.