A framework of view-dependent planar scene active camouflage

Active camouflage is a technique for occluding objects disappear from the observer. Current implementations make assumptions on the viewpoints of both the observer and the camouflaged target. In this work, we present a framework of view-dependent planar scene camouflage. The cameras representing the observer and the occluding object are placed in general positions. The images captured by the object camera are transformed to the observer viewpoint to generate globally consistent visual data for transparent camouflage. Experimental results are presented for real scene images.