Tomography is commonly used to build the anisotropic model for depth imaging, which requires human intervention for residual picking and quality control. In addition, tomography is typically implemented with a gradient descent optimization that can lead to a local extremum. We present an automatic inversion approach to estimating Thomsen parameters ? and ? when vertical velocity is known from well data with the assumption of 1-D vertical transverse isotropy (VTI).It employs the very fast simulated annealing as the global optimization algorithm and a fast model based moveout technique for re-migration/de-migration engine. We demonstrate the automatic inversion with both synthetic and field data.