Multichannel Reflectivity Inversion with Sparse Group Regularization Based on HPPSG Algorithm

Sparse -Spike Deconvolution (SSD) is commonly used in seismic deconvolution. However, when dealing with multichannel seismic data through the trace by trace procession, it can't maintain the lateral continuity and stability well. In this paper, we propose a new SSD method based on Hadamard Product Parametrization Sparse-Group algorithm (HPPSG). In order to preserve lateral continuity, we define each layer of the seismic profiles as a group, and then use L_p regularization (1≤p) to constrain each element in this group. Assuming that reflectivity is sparse, we apply L_q (q≤1) as a regularization to constrain between groups along the time direction. Then, we construct a L_(p,q) optimization problem. After that, we solve this problem using HPPSG algorithm based on the Hadamard Product Parametrization Lasso algorithm (HPPL). Synthetic and real data examples indicate that the proposed method have significant improvements on the lateral continuity.