Boundary Surface Shrinking - a Continuous Approach to 3D Center Line Extraction

Abstract

We present a new algorithm for efficient and robust approximation of skeletons and construction of connected center line graphs from 3D image data. The algorithm is based on the idea of shrinking the boundary surface along the gradients of the object's distance map. After transforming the surface in that way, duplicate vertices and line segments are eliminated. If the skeleton contains no medial surfaces, a graph representation of the centerline can directly be extracted from the remaining edges. The algorithm has proved to perform well on several different datasets. 1 Introduction Much work has been done in the field of skeletonbased shape analysis. One application of increasing interest is the extraction of center line graphs from skeletons in order to represent branching structures like for example neurons with their axons, dendrites, and spines. Existing 3D skeletonization methods suffer from several problems, the most severe ones being the discrepancy between discrete and continu...