Virtual manipulation of hidden objects

An occluded object displayed on a head-mounted display (HMD) is manipulated by establishing a first view position relative to a real-world object, and receiving three-dimensional data associated with occluded objects, the occluded objects at least partially occluded by the real-world object at the first view position. A first image of at least a portion of the occluded objects is rendered on the HMD from the first view position at a geometrically correct location with respect to an image of the real-world object on the HMD using the three-dimensional data. At least one occluded object is selected from the rendered portion of the occluded objects, a second view position is established relative to the selected at least one occluded object, and a second image of the selected at least one occluded object is rendered on the HMD from the second view position using the three-dimensional data.