Any un-modeled object in the scene will not be detected by ordinary occlusion detection algorithms like the ML3DImage algorithm. Such algorithms detect the occlusions based on the geometry while the un-modeled objects appear only in photos. Therefore both the objects geometry and the objects texture are employed in the Photo Occlusions Finder (POF) algorithm.

The main characteristics of this algorithm are the ability to detect various types of occlusions (modeled/un-modeled) and the flexibility to deal with different photos captured by different sensors. The reliability of the developed algorithms is then demonstrated through some practical applications.