Winston Lewis

Regularization And Full-waveform Inversion: A Two-step Approach

Geophysical inverse problems are usually ill-posed and illconditioned, meaning that their solution is not unique and/or unstable. This kind of inverse problems can be solved using regularization techniques from optimization theory, where an additive or multiplicative regularizer will help obtain an automatic selection of a unique and stable solution for a given cost function. Regularization helps condition an inverse problem by providing the search for possible solutions with a priori information that comes from inferences about the physics of the problem. There is a toolbox of regularization techniques for inverse problems. Thus, it is necessary to assess the effect of different regularizers and different a priori information to select the best combination for the problem at hand. In this expanded abstract, we focus on regularization techniques aimed at stabilizing and improving the model estimated through full waveform inversion by adopting a two-step approach in light of the computational challenges posed by inverting for large 3D models.