May 2, 2017
Jolley Hall, Room 309
"Medial Axis Computation and Application"
Adviser: Tao Ju
Medial axis is a classical shape descriptor. It captures both the geometry and topology of a shape compactly. Therefore medial axis is a useful tool for analyzing shapes in several fields, including biology, computer vision and computer graphics. Though its value is well appreciated, robust computation has been lacking because medial axis is sensitive to noise on the shape's boundary and difficult to approximate. We contribute to the community by tackling these two challenges. First, we define a novel significance measure to de-noise the medial axis of a 3D shape. Second, we propose a scalable and accurate method for approximating the medial axis of an important class of shapes, the digital shapes. Finally, we argue that medial axis has the potential to help solve real world problems. Particularly, we propose to recover the correct structure of a complex crop root system using medial axis.