Merging of overlapping two-dimensional (2D) images that are formed by an image pick-up device as projections of a three-dimensional (3D) scene includes image registration by projective transformation of one of the 2D images. The transformation is derived from corresponding feature points found in both images. In order to achieve improved accuracy and stability, the coordinates of the corresponding feature points are translated so that, on average, the numerical ranges of coordinate values are minimized. An apparatus for performing the merging of overlapping 2D images includes an appropriately configured image processor or computer with an attached image acquisition device. In one embodiment, the apparatus is a diagnostic x-ray apparatus.