In this paper, we intend to provide a method for calibration of geometric CBCT systems without the need for phantom and prior knowledge related to the object which is normally used in 2D-3D registration methods for geometrical calibration. To this end, we use an iterative algorithm based on comparison between measured data and DRRs1 which are obtained from reconstructed images with different geometries to find geometrical parameters of the system. Our proposed method shows strong agreement with phantom based methods.