Contour Matching Using Epipolar Geometry

Generally contour matching is diicult, and the problem becomes more diicult when the motion between successive image frames is large. When the image frames are obtained while the camera is in motion, we can use constraints about the scene and the camera. In this paper we propose a contour matching algorithm which guarantees an accurate matching result, even for large motion contours. The key idea is to use epipolar geometry and geometric constraints. This contour matching method can be applied to contours from two or three diierent views. All of the processes are fully automatic and are successfully implemented and tested with many images.