In order to simulate the actual geological conditions, the authors present the least squares inversion by incorporating topography into a forward model. In consideration of an ill-posed inverse problem of MT, the authors introduce Tikhonov regularization to obtain the equation of the total objective function and utilize smoothness-constrained least-squares inversion to solve the total objective function. As the regularized factor controls resolution and stability of the inverse problem, the authors put forward active constraint balancing (ACB) to obtain an optimized regularized factor that balances the resolution as well as the stability of the inversion process. Meanwhile, for the purpose of speeding up the calculation of the field Jacobian for 2-D magnetotelluric inversion, the principle of electromagnetic reciprocity is applied. Finally, the authors discuss the inversion results of TE mode, TM mode and joint inversion of TE and TM mode using some synthetic models, in comparison with some of previous work.