This work was supported partly by a Bridgestone Corporation fellowship and partly by the Army Night Vision and Electro-Optics Laboratory with contract DAAK20-85-C-0293


This research is motivated by a desire to integrate some of the diverse, yet complimentary, developments that have taken place during the past few years in the area of passive stereo vision. On the one hand, we have approaches based on matching zero-crossings along epipolar lines, and, on the other, people have proposed techniques that match directly higher level percepts, such as line elements and other geometrical forms. Our rule-based program is a modest attempt at integrating these different approaches into a single program. Such integration was made necessary by the fact that no single method by itself appears capable of generating usable range maps of a scene.

Date of this Version