These approaches exploit the fact that the invisible
background part in the left view may be visible in the right
view [14], [15], and then the disoccluded region in the virtual
view warped from the left view can be filled with the background
information from the right view.