Sparse seismic inversion is an effective strategy to restrict the sparsity of reflection coefficients. Considering the excellent anti-noise property of sparse inversion, we proposed one novel mix-domain sparse inversion utilizing fast matching pursuit algorithm. In our work, the mixed frequency-domain model introduced by Margrave (1998) is regarded as the forward equation. Firstly, the object function is constructed with fast matching pursuit algorithm. And the low-frequency constraint is also introduced into inversion process to enhance the stability of inversion algorithm. Conventional matching pursuit inversion can only obtain one optimal atom after an iteration which can achieve the maximum absolute value in inner product space. However, fast multi-atom matching algorithm obtains more than one atoms which can improve the computation speed and convergence precision effectively. In addition, we illustrate the feasibility and excellent stability through several synthetic tests. Finally, one field case is studied carefully using time-domain Bayesian inversion and our method. Superior inversion results with our method can also maintain a higher fitting degree to the real logging data.
展开▼
