3D whole-heart coronary MR angiography (CMRA) acquisition remains lengthy and can suffer from residual motion and/or undersampling related artifacts. 2D image-navigator based non-rigid respiratory motion compensation has been recently proposed to accelerate the CMRA scan. This framework combines 2D beat-to-beat translational and 3D bin-to-bin non-rigid motion correction. However, beat-to-beat anterior-posterior motion is not corrected for with this approach, which can result in significant residual motion. Here we propose a virtual 3D iNAV approach that exploits autofocus motion correction to further enable beat-to-beat anterior-posterior translational motion correction, assuming a linear relationship between the translational foot-head and anterior-posterior movement of the heart.