Least-squares reverse time migration (LSRTM) is a leading tool in imaging complex geological structures. This is because it includes the full-wave equation via reverse time migration (RTM), and solves the linearized wave equation by the least-squares (LS) minimization. To stabilize the LSRTM solution and mitigate the shortcomings that arise from linearization, simplification, approximation in simulation, and imperfection of data acquisition an appropriate regularization must be used. However, the conventional regularization methods such as Tikhonov, total variation (TV), or sparse regularization perform suboptimally, even in recovering structurally simple earth models. Here we introduce a new extension of Tikhonov-TV compound regularization, called shifted TV, to regularize the unknown reflectivity image. This new regularization implemented in the frame of a nonlinear migration allows for high-resolution and stable imaging in the presence of rough migration velocity models. We demonstrate the performance of our method using a short offset academic-style streamer data generated in the crustal-scale GO 3D OBS synthetic subduction zone model. The results confirm that the shifted regularization increases the robustness of LSRTM with an imperfect background velocity model and allows us to estimate high-quality reflectivity images of complex geological setting despite the limited streamer length.
Note: This paper was accepted into the Technical Program but was not presented at IMAGE 2022 in Houston, Texas.
