Enhancement of Fast Acquired Disparity Maps Using a 1-d Cellular Automaton Filter

Among others, stereo vision analysis deals with the extraction of depth in a scene, using the disparity between the images acquired by a stereo camera setup. The disparity calculation for the whole of an image is mostly a computation demanding procedure, commonly being performed by dedicated hardware. In this paper a hardware architecture for real time extraction of disparity maps is proposed, capable of processing images of 1MPixels in less than 25ms. The produced disparity maps are indented to be used for real-time navigation of a mobile platform, as well as in other time critical applications which demand 3D information.