An automated registration method for range images

Abstract

Registering range images captured by 3D scanner and constructing a complete model of an object is frequently used in reverse engineering, computer game and animation, digital museum and digital heritage, etc. Many algorithms about registering the range images have been developed up to now. In this paper, a robust automated algorithm is presented based on surface characteristics extracting and matching techniques. The algorithm restores a set of point correspondences between every two adjacent range images without image position information or user interaction. Differential quantities are used as geometric descriptor and some other constraints are used to accelerate searching and matching processes and maintain result accuracy. After calculating the estimate of the rigid motion transformation between range images with selected point correspondences, a modified ICP algorithm is used to refine the result. The whole algorithm can be applied in many areas, especially in building the digital models of cultural artifacts and works of art. © Springer-Verlag Berlin Heidelberg 2007.