Authors:
Jun Zhang
1
;
Zhouhui Lian
2
;
Zhenbao Liu
3
and
Jianguo Xiao
2
Affiliations:
1
Peking University and Northwestern Polytechnical University, China
;
2
Peking University, China
;
3
Northwestern Polytechnical University, China
Keyword(s):
Mesh Segmentation, Shape Descriptor, Convexity, Fast Marching, Local Depth.
Related
Ontology
Subjects/Areas/Topics:
Computer Vision, Visualization and Computer Graphics
;
Geometric Computing
;
Geometry and Modeling
;
Modeling and Algorithms
;
Scene and Object Modeling
;
Surface Modeling
;
Texture Models, Analysis, and Synthesis
Abstract:
Mesh segmentation is a fundamental way of shape analysis and understanding for 3D mesh models. In this paper, we propose an effective heuristic mesh segmentation algorithm, which is based on concave areas detection and heuristic 2-category classification via fast marching. The algorithm has several merits. First, the boundary between each pair of segments is close to the natural seams of 3D objects. Second, it is robust against pose variations and isometric transformations. Finally, our algorithm decomposes non-rigid 3D models into a set of rigid components in a short period of time and the procedure is fully automatic. Extensive experiments in this paper demonstrate that the proposed method outperforms the state of the art in mesh segmentation.