Diffuse axonal injury (DAI) is a common pathology in traumatic brain injury (TBI), which occurs when sudden accelerations to the head cause excessive stretching and shearing of axonal fibers, eventually leading to neuropathologic damage and dysfunction. To accurately capture the full-field kinematics of brain and its substructures under such large deformations and high rates of loading, a fully subject-specific three-dimensional computational human head model is developed in this study. Specifically, the model geometry is based on the brain anatomy and morphology obtained directly via magnetic resonance imaging (MRI), the material properties of individual brain regions are derived from in-vivo magnetic resonance elastography (MRE) at three actuation frequencies (30, 50 and 70 Hz), and the brain kinematics data under mild rotational accelerations to the skull, which is used for model validation, is extracted from tagged MRI imaging technique. Notably, unlike the currently available computational models that employ linear/quasi-linear visco-hyperelastic constitutive models and thus exhibit poor accuracy beyond the small strain regime (~0.1%), the computational model developed in this work employs a recently proposed viscous dissipation-based finite strain visco-hyperelastic constitutive model, which can effectively capture the nonlinear strain stiffening and strain rate sensitivity responses that are characteristic of brain tissue under injury-relevant loading conditions. To derive the material parameters of this model, a hybrid in vivo-ex vivo parameterization method is developed, in which an initial optimal parameter set is first calibrated from the linear viscoelastic material response obtained from MRE, which is then improved via a nonlinear constraint optimization to fit the large strain viscous overstress from ex vivo mechanical testing data in the literature. The resulting head model simulations show reasonable agreement with the tagged MRI validation data.