Octree encoding of objects from range images

摘要

An efficient algorithm for the construction of an octree model from range images is proposed. A 3D model constructed from intensity images cannot describe the free surface of an object accurately since this construction only makes use of the silhouettes of a 2D image. A 3D model constructed by the method proposed here, however, can describe the free surface, since this method makes use of both silhouette and depth information of a range image as well as the silhouettes of intensity images. It is shown that the proposed algorithm is less restrictive than existing methods for building a 3D model from arbitrary views.