The invention relates to the registration of three-dimensional image data of an imaged object, in particular of pre-operatively taken 3D image data of a bone structure of a patient, with a set of two-dimensional projection images of the object.The three-dimensional image data of the imaged object (16) comprise at least a first rigid part (21a) and a second rigid part (21b), wherein the first and the second rigid parts are movably connected to each other. The three-dimensional image data represent a region of the imaged object (16) containing the first rigid part (21a) and the second rigid part while the first and the second rigid parts are in a first state of position and orientation relative to each other. Sub-regions corresponding to the first and second rigid parts within the three-dimensional image data are identified and the three-dimensional image data are divided into at least first image data of a first sub-region representing the first rigid part (21a) and second image data of a second sub-region representing the second rigid part. A set of two-dimensional projection images of the imaged object (16) (and e.g. taken by imaging device 14) comprises at least one image taken in a first projection direction and at least one image taken in a second projection direction, the first and second projection directions being different projection directions. The set of two-dimensional projection images represents a region of the imaged object (16) containing the first rigid part (21a) and the second rigid part while the first and the second rigid parts are in a second state of position and orientation relative to each other. A processing device (19) registers the first image data of the first sub-region with the set of two-dimensional projection images thereby obtaining first registration information, and separately registers the second image data of the second sub-region with the set of two-dimensional projection images thereby obtaining second registration information.
