Robust Estimation of Multiple Local Dips via Multidirectional Component Analysis

We propose a novel method for robust estimation of multiple local dips in seismic data via signal decomposition. The decomposition is achieved by regularizing each separated component to have a single dip. The single-dip regularization is built based on the fact that piecewise constant dips indicate that local gradient vectors lie on the same directions. The method is formulated as an optimization problem, which is solved in the framework of a block coordinate descent method. The proposed method is robust to random noise and can be used to estimate multiple dips accurately. As the proposed method has clear physical meaning, it is easy to be applied for other tasks, such as random and coherent noise attenuation, signal separation, and so on. We test the feasibility of our method in estimating local dips with synthetic and field data and compare with the structural tensor method and the plane-wave destruction method. Results demonstrated that the proposed method can estimate multiple dips and is more robust to noise than the other two methods. Applications for random noise attenuation, coherent signals separation, and unconformity detection are also tested.