Skeleton Graph Generation for Feature Shape Description

Abstract

An essential step in feature extraction is the calculation of attribute sets describing the characteristics of a feature. Often, attribute sets include the position, size, and orientation of the feature. These attributes are very important, but they do not provide a good approximation of the shape of a feature. For better shape description, a more sophisticated method is needed. This paper describes a method that extracts a binary skeleton of a feature, and transforms it into a graphical representation: the skeleton-graph. This graph represents the original skeleton with controlled precision, and contains the essential topology and geometry of the skeleton. In addition, distance information is used to generate a simplified reconstruction of the original 3D feature shape, which can also be used as an iconic object for visualization.