Full Waveform Inversion with Optical Flow

Abstract

Full waveform inversion (FWI) is based on updating process by a gradient direction which is dependent on the difference between the observed and modelled data. However, the gradient direction of the misfit function suffers from local minima due to errors in the data residual and starting velocity model, then FWI derives the solution which is no longer the optimum ones. In this paper, we present a new approach to decompose wavefields into angle domain using optical flow. Having direction of the wavefields by optical flow, it allows to compute the reflection angle between both forward and backward wavefields. This new FWI algorithm can use large reflection angles help retrieving long wavelength structures and attempt to reduce uncertainties of falling into local minima, and minimize cycle skipping and migration footprint in the gradient direction. Thus, the application of optical flow in FWI can play a role that makes FWI be more robust and stable by the angle-dependent gradient computation.