Item

Abstract

We present a technique for recovering structural information from triangular meshes that can then be used for segmentation, e.g. in reverse engineering applications. In a preprocessing step, we detect feature regions on the surface by classifying the vertices according to some discrete curvature measure. Then we apply a skeletonization algorithm for extracting feature lines from these regions. To achieve this, we generalize the concept of morphological operators to unorganized triangle meshes, providing techniques for noise reduction on the binary feature classification and for skeletonization. The necessary operations are easy to implement, robust, and can be executed efficiently on a mesh data structure.