I. Introduction
2D/3D overlay and cone beam computed tomography (CBCT) are state-of-the-art applications for nowadays C-arm systems. The spatial accuracy within the reconstructed volume which is required for CBCTs is up to and up to for 2D/3D overlays for stent placing [1], [2]. Without a calibration these accuracies are not attainable [3]. The common way to improve the spatial accuracy of C-arm systems is a data-based approach. A high repeatability of the system is assumed, which means for two acquisition from the identical C-arm system pose the X-ray projection geometry stays the same. For the calibration, an X-ray of a phantom with known geometry is performed. Subsequently, the projection geometry is estimated by an optimizer and stored. Afterwards, the estimated projection geometry can be taken for each acquisition which was calibrated before, such that a 2D/3D overlay or a CBCT is feasible [4]–[6]. An interpolation or extrapolation in Cartesian space for the use of other system poses is possible, however, errors with unknown amount may appear, because the system is positioned in joint space.