Meaningful Mesh Segmentation Based On Saliency Analysis

3D computer models are increasingly popular due to the advent of the 3D laser scanning system and the boom of Virtual Reality applications. These models are often represented as complex polyhedral meshes, which may have thousands or even millions of vertices and polygons. Although this representation is useful for visualization, a polygonal mesh does not capture high-level structure. In general, high-level abstractions are useful for managing data in applications, such as object registration, object retrieval and indexing, feature modeling, mesh optimization, etc. One way to impose such high-level description is through mesh segmentation.This thesis, in both the theoretical perspective and the practical perspective, probes into 3D model segmentation. A survey was given to summarize 3D mesh model segmentation techniques, including latest achievements, classification and application in this field. The main contributions of this thesis are summarized as follows:(1) A novel algorithm for 3D mesh segmentation based on saliency analysis is proposed. We segment the mesh considering human visual attention. Most segmentation methods decompose 3D objects into parts based on curvature analysis. But this algorithm computes the mesh saliency firstly, and then uses the vertices that have the maximum mesh saliency as the boundary. Because it's known that repeated patterns, even if high in curvature, are visually monotonous. It is the unusual or unexpected that delights and interests. So the difference between most of the former mesh segmentation algorithm is not to use the vertices of highly negative curvature as boundary, but to use the vertices have the maximum mesh saliency. We then use a fast marching watershed scheme. It is based on the classic watershed algorithm but cost less.(2) The typical 2D silhouette parsing short-cut rule is introduced to 3D mesh segmentation domain. Based on the extended 3D short-cut rule theory, a hierarchical shape decomposition paradigm is proposed, which integrates the advantages of skeleton-based and minima-rule-based meaningful segmentation algorithms. The method makes use of new geometrical and topological functions of skeleton to define initial cutting critical points by sweeping a given mesh perpendicular to every skeleton branch, and then employs salient contours with negative minimal principal curvature values to determine natural final boundary curves among parts. And sufficient experiments have been carried out on many meshes, and shown that our algorithm can provide more reasonable perceptual results in more robust way.