A 3D result image of an object is reconstructed from a set of X-ray two-dimensional projections of the object. A 3D reference image of the object is reconstructed by employing a compressed sensing technique based on at least some of the 2D projections at a reference motion state of the object. By employing an algebraic and/or analytic reconstruction technique, 3D intermediate images are reconstructed for various motion states of the object. The 3D intermediate images are registered with the 3D reference image to obtain spatial transformations for the various motion states of the object. Based on the spatial transformations, the 3D intermediate images are transformed to a joint phase and combined to obtain the 3D result image.
