Range Image Segmentation Through Fuzzy and Possibilistic Shell Clustering

Abstract

Segmentation and surface approximation (i.e., description of surfaces in terms of parametrized surface patches) is an important step in range image analysis. In this paper, we show that the surface approximation problem can be formulated in terms of a C-means type fuzzy clustering algorithm with suitable (shell-like) prototypes and distance measures. We present an unsupervised fuzzy clustering algorithm to obtain an approximation of a range image in terms of planar and quadric surface patches.