A Generalized Tensor-based Coherence Attribute

Abstract

The coherence attribute is one of the most popularly used seismic attributes. It has been widely used for highlighting structural and stratigraphic discontinuities such as faults, fractures, and channels. We show that the popular C3 coherence algorithm is a special case of a Generalized Tensor-based Coherence (GTC) attribute. This attribute is computed by unfolding a seismic analysis tensor along different modes corresponding to time, inline, and crossline directions. We also present a novel preprocessing method based on a multivariate Gaussian kernel for computing coherence on seismic volumes. This method weighs seismic traces based on their relative proximity to the reference trace. Thus significantly reducing the noise in the covariance matrices, and amplifying the features close to the reference trace. We show that the proposed GTC coherence and the preprocessing method result in superior detail compared to other coherence attributes, and gives more freedom and flexibility to the seismic interpreter.