PREPARING THE INITIAL MODEL FOR ITERATIVE DEBLENDING BY MEDIAN FILTERING

摘要

Simultaneous-source acquisition can obtain much faster acquisition with higher spatial sampling at the cost of obtaining highly noisy seismic records. While researchers are seeking specific direct imaging algorithms for attenuating the simultaneous-source interference during the migration process, deblending is still a preferable way to deal with simultaneous-source seismic data at the current stage. Removing blending noise while preserving as much useful signal as possible is thought to be the key in the deblending process. Previous study has shown that the median filtering (MF) can be used to effectively and efficiently separate the simultaneous-source seismic data in the common midpoint domain based on one-step filtering. In this paper, we investigate the application of MF in preparing the initial model in an inversion-based deblending framework. Our results show that the application of MF in a shaping regularized iterative deblending framework can effectively accelerate the convergence rate and improve the deblending performance in a limited number of iterations. We use three different synthetic examples and one field data example to demonstrate our proposed methodology.