| Article ID | Journal | Published Year | Pages | File Type |
|---|---|---|---|---|
| 441989 | Computers & Graphics | 2013 | 12 Pages |
•Grouping sets of scalar functions according to their behaviour over a shape, or a shape class.•Selection of functions that are as much as possible selection independent.•Use of a clustering technique to learn representative functions.
Scalar functions are widely used to support shape analysis and description. Their role is to sift the most significant shape information and to discard the irrelevant one, acting as a filter for the characteristics that will contribute to the description. Unfortunately, a single property, or function, is not sufficient to characterize a shape and there is not a method to automatically select the functions that better describe a 3D object. Given a set of scalar functions defined on the same object, in this paper we propose a practical approach to automatically group these functions and select a subset of functions that are as much as possible independent of each other. Experiments are exhibited for several datasets to show the suitability of the method to improve and simplify shape analysis and classification issues.
Graphical abstractFigure optionsDownload full-size imageDownload high-quality image (249 K)Download as PowerPoint slide
