Plane Detection in Disparity Space

We report an algorithm for detecting planes from disparity maps, without computing 3-D information explicitly. Camera calibration is not assumed unless Euclidean plane parameters are required. The algorithm identifies co-planar image patches, thus providing a useful tools in the presence of occlusions. The basic components are a closed-form expression for the disparity of a plane, and a Hough-like voting scheme for plane identification. Experimental results with real data suggest promising performance, comparable with that of recently reported plane detection methods.