A method for building multiview models of curve surface objects

Summary form only given. A method has been developed for building multiview models of a curved surface by using direct input images from a camera, representing the curved contour with polygon approximation, and then taking the topological and region information of each project image as features of these models. The monocular camera is used to represent the relation of the camera focal length to the distance between the camera and the object in order to adapt for the change of both parameters. The goal is to simplify the representation of a curved object and the matching processing of models with images. A set of images of curved surfaces taken from different view points is input and processed to obtain the set of contour images. The contour images are converted into polygons according to the variance of the contour's curvature. The topological and region information are extracted from these polygon images and the multiview models of an object are built from them.<
>