New attributes for salt dome detection in 3D seismic data using higher order SVD

Abstract

In this paper, we propose new attributes for seismic interpretation in 3D volumes using Higher Order Singular Value Decomposition (HOSVD). HOSVD has the advantage of taking into account multidimensional data and therefore can effectively be used for 3D seismic data. Contrary to many existing attributes that are extracted from 2D windows, the proposed algorithm takes into account the continuity of medium and uses a seismic volume to compute the trace, largest singular value, and coherence attributes. These attributes can be used for detecting salt domes, horizons, chaotic horizons, and faults. In this work, we focus mainly on the problem of salt dome detection in 3D seismic data. The proposed workflow uses the estimated attributes from the time, crossline and inline directions. A combination of these attributes ensures that the proposed algorithm works well even when the salt boundary is represented by weak reflectors. We have used the publicly available Netherlands offshore F3 dataset to evaluate the performance of the proposed salt dome detection algorithm. Our experimental results show that the proposed method using the new attributes can detect salt boundaries with superior accuracy outperforming other edge and texture based methods.