Registering range images and segmenting 3d object models

We propose a crude range image registration algorithm based on edge detectionin spherical coordinates. We used the edge features obtained by our edge detectionmethod to register successive range images instead of using the whole point infor-mation. Different from previous edge detection methods, we first obtain a functionrepresentation in spherical coordinates. This representation allows detecting smoothedges on the object surface easily by a zero crossing edge detector. We use the wellknown ICP algorithm on these edges to register patches in a crude manner. Then, weapply the ICP to whole point set and obtain the final registration. The dual operationperformed extremely fast compared to directly registering the points sets without anyinitial estimation. Our method also produced lower registration errors compared to theconventional ICP algorithm. One other important byproduct of our method is that,we can obtain the edges of the 3D object while registering it. These edge points maybe of use in 3D object recognition and classification.We also provide a free-form 3D object segmentation algorithm based on the sameedge detection algorithm in spherical coordinates. Physically meaningful segments of a3D object may be useful in various computer graphics applications. In order to segmentfree-form objects, we introduced a novel method in this study.