Perpustakaan Digital ITB

With ill-posed inverse problems such as AVO inversion, regularization schemes are required to obtain stable and unique solutions. The total variation (TV) regularization method is commonly used to resolve sharp interfaces and obtain blocky solutions, where edges and discontinuities are preserved. TV regularization accomplishes these goals by imposing sparsity on the first-order difference (gradient) of the model parameters. In this paper, we first explore the use of TV regularization in the three-term AVO inversion problem. Then, we illustrate that the discretization and sampling in data processing may yield complex shapes of models, whereas conventional TV regularization is not applicable to such situation. Therefore, we adaptively refine the original TV regularization by updating the order of difference operator and propose a high-order TV regularization scheme for non-block-structured models. The method can effectively use the structural information of the models, which ultimately results in better-quality solutions. The rationality of the approach is also confirmed by using sparse theory and a reconstruction algorithm. Moreover, to solve the inversion problem, we improve an existing optimization algorithm via a step search process to obtain higher efficiency and lower residual error. Tests on both synthetic and real data show that our method can provide a credible and high-resolution estimation of the subsurface models.