Stacking Using Truncated Singular Value Decomposition and Local Similarity

J.Y Xie,B.R Di,J.X Wei,Q. Xie,S.H Zu,Y.K Chen

DOI: https://doi.org/10.3997/2214-4609.201601325

2016-01-01

Abstract:Summary The similarity-weighted stacking takes use of the local similarity between each trace and a reference as the weight to stack the NMO-corrected prestack seismic data. The selection of reference trace plays a significant role in the final performance. The traditional similarity-weighted stacking approach uses the traditionally stacked trace as the reference, which is not plausible when there is abnormal trace in the data. We proposed a new criteria for selecting the reference trace, which is based on the stacked trace after applying a truncated singular value decomposition to the NMO corrected data. The TSVD aims to reduce the impact of the abnormal trace in the stacked trace, and thus can make the similarity-weighted stacking more robust. We use both synthetic and field data examples to show the successful performance of the proposed approach.