A new approach to coherency inversion by global optimization

Global coherency inversion, which consists of maximizing a semblance function calculated from prestack seismic data, has been proposed for estimating the velocity-depth model with no necessity for prestack or poststack travel time picking. At each iteration of inversion, multi-offset travel time calculation is performed and the semblance value is computed on prestack gathers along these time curves. Layer velocities and zero-offset times for principle reflectors are chosen as inversion parameters. In this work, coherency optimization is based on a new Regularized Global Approximation method (unlike simulated annealing or genetic algorithms, it includes a predictor-corrector procedure with significantly reduces the number of evaluations of the objective function.