Feature-Preserving Medial Axis Noise Removal

Roger C. Tam, Wolfgang Heidrich

Abstract: This paper presents a novel technique for medial axis noise removal. The method introduced removes the branches generated by noise on an object’s boundary without losing the fine features that are often altered or destroyed by current pruning methods. The algorithm consists of an intuitive threshold-based pruning process, followed by an automatic feature reconstruction phase that effectively recovers lost details without reintroducing noise. The result is a technique that is robust and easy to use. Tests show that the method works well on a variety of objects with significant differences in shape complexity, topology and noise characteristics.