oa Migration Velocity Analysis with Maximum Stack Image Power

Waveform inversion can be implemented in either data domain or in image domain, which is usually called migration velocity analysis. As a data-domain approach, classic waveform inversion attempts to find a velocity model that minimizes the misfit between the real data and the simulated data. This kind of inversion has little success for the past decades. We believe that one of the reasons is the unknown wave equations, which is used to simulate the wavefields. Therefore, an effective waveform inversion should emphasize on the events' traveltime or phase information and downplay the role of amplitude information. Based on the criteria that images must be geometrically coherent after migration, we propose an image-domain approach. When the velocity is correct, the events at CIGs have maximum stack power for redundant shots. This approach uses RTM as the migration engine. Our optimization problem is therefore correspondingly defined as a semblance-like objective function, which is constrained by RTM's forward and backward two-way scalar wave equations. The gradient of the objective function is derived by adjoint state technique.