Computing surface information from monocular and binocular cues for vision applications

Surface perception involves computing shape information from several geometric cues. An underlying problem confronted here is to build a scheme by which information from different cues could be combined to achieve a unified percept. The authors address the issues involved in integrating the visual cues. The need for using the stage of stereo data interpolation as the appropriate level in the visual hierarchy is argued. With a suitable modification in the interpolation scheme of D. Shephard (1964), integration of binocular disparity and monocular cues is shown to be accomplished.<>