An efficient implementation of eigenstructure-based coherence algorithm using recursion strategies and the power method

The eigenstructure-based coherence algorithm is an appropriate method for imaging faults and stratigraphic features. It produces coherence images with high resolution and has good anti-noise ability. However, the large computational cost of constructing the covariance matrix and calculating the dominant eigenvalue restricts its application, and earlier researches have not discussed how to sustain the computational efficiency of this algorithm. The focus of this paper is the efficiency problem, and we propose the use of recursion strategies and the use of a specific power method to accelerate the calculation of the eigenstructure-based coherence algorithm. We use repetitions of the spatial correlations in the covariance matrices of neighbouring points to reduce the computational cost of constructing the covariance matrices. Then, we use the power method to calculate the dominant eigenvalue of the covariance matrix and choose the eigenvector associated with the dominant value at time t as the initial vector at time t + 1. The results on field data show that our implementations can greatly reduce the calculation time of the coherence attribute.

► The cost of C3 algorithm is on constructing covariance matrix and eigendecomposition.
► Earlier paper ignored to discuss how to keep this algorithm computation efficient.
► The recursion is used to reduce the cost of constructing matrix.
► We choose the initial vector of power method to reduce the computation cost.
► The results on field data show the computational cost can be greatly reduced.