HIGH-ORDER SPARSE RADON TRANSFORM FOR DEBLENDING OF SIMULTANEOUS SOURCE SEISMIC DATA

摘要

This paper proposes an iterative high-order Radon transform based on matching pursuit (MP) algorithm to separate the blended seismic data. During each iteration, the matched subspace is picked by energy distribution in the high-order Radon domain. In thus small subspace, the high-order Radon transform is realized quickly to estimate the effective signals. The blending noise is then estimated by the estimate-deblended data with prior acquisition code and subtracted from the pseudo-deblended data. Thus an iteration is finished. The MP method shows more sparse than the iterative reweight least square method (IRLS). We compared the denoising effectiveness between these two methods. Synthetic and field data experiments prove that the matching pursuit algorithm has higher SNR and better denoising effectiveness than IRLS method.