GII regularization technique for seismic data inversion

This work proposes a new regularization strategy for the Full Waveform inversion (FWI), a promising technique to generate images of the subsurface from seismic data. This problem is ill-posed and depends on the application of regularization strategies to minimize noise fitting. The most often used strategies are Total Variation and Tikhonov regularization, assuming that the solution has a certain degree of smoothing. The main objective of this paper is to propose the use of the Geological Incoherence Index - GII as a constraint for the inverse problem. The GII assumes that the values of velocities increase with depth, perpendicularly to the interfaces between layers. The main advantage of the GII regularization is that it does not change the search direction of the FWI without regularization. In the present work, the forward problem is modeled by the Helmholtz equation, and solved by the Finite Difference Method. The inverse problem is solved by a multi-scale strategy using the L-BFGS method. Computational experiments presented in this paper show that the use of GII leads the solution to a more geologically likely solution, providing more realistic images of the subsurface compared to the classic FWI, since it carries information of the sedimentary bedding. (C) 2018 Published by Elsevier B.V.