Extracting Feature Curves on Point Sets

Abstract

We present an effective algorithm for detecting feature curves on point sets. Based on the local surface fitting method, our algorithm first compute the curvatures and principal directions of each point of point sets. The algorithm then extracts potential feature points according to the biggist principal curvature of the point, and evaluates the principal directions of the detected points. By projecting the points onto the principal axes of their neighborhoods, the potential feature points are smoothed. Using the principal directions with each optimized point, feature curves are generated by polyline growing along the principal directions of feature points. The results indicate that our algorithm is sensitive to both sharp and smooth feature curves of point set, and it supports multi-resolution extraction of features.