Tomographic reconstruction requires precise knowledge of the position of the center of rotation in the sinogram data; otherwise, artifacts are introduced into the reconstruction. In parallel-beam microtomography, where resolution in the 1 microm range is reached, the center of rotation is often only known with insufficient accuracy. We present three image metrics for the scoring of tomographic reconstructions and an iterative procedure for the determination of the position of the optimum center of rotation. The metrics are applied to model systems as well as to microtomography data from a synchrotron radiation source. The center of rotation is determined using the image metrics and compared with the results obtained by the center-of-mass method and by image registration. It is found that the image metrics make it possible to determine the axis position reliably at well below the resolution of one detector bin in an automated procedure.