Feature-Sensitive Mesh Simplification

Feature recognition in meshes is a widely studied topic of computer science and is applied in order to identify and analyse properties of a 3D shape.It finds applications in numerous fields, like, e.g.medical imaging, robotics, cultural heritage, and CAD design.In this paper, we propose a new segmentation-based feature recognition approach for 3D meshes, where features are defined by the intersection of two neighboring segments.The detected features are used to derive very sparse polygonal representations of arbitrary input meshes, even merging features to achieve a high degree of simplification.We compare results for edge-based segmentation and a new normal-based segmentation and demonstrate how to derive good results even for pseudocurved surfaces and meshes affected by artifacts.