A fast algorithm to estimate the dominant eigenvalue of Real Symmetric Matrices and its application in C3 algorithm

The third-generation coherence algorithm, C3, is robust to noise and posseses high resolution. However, the process of computing eigenvalues of the covariance matrix in C3 is time consuming. To avoid computing all eigenvalues, this paper proposes a fast convergence algorithm based an eigenvalue estimation of real symmetric matrices to calculate the dominant eigenvalue. To control the precision of the algorithm, an error evaluation formula is presented. By adding traces along the boundary of the seismic data volume, we avoid judging whether the spatial window overstep the boundary when apply recursion strategy in a horizontal direction. Application to real data shows efficiency of C3 is improved by approxinately 3 times.