Voronoi-based estimation of Minkowski tensors from digital images

CSGB seminar
Thursday, 11 September, 2014, at 14:15-15:00, in Koll. G3 (1532-218)
Anne Marie Svane (Department of Mathematics, Aarhus University)
Abstract:

Minkowski tensors are used in shape analysis to quantify such properties as position, orientation and eccentricity of an object. Other important special cases are the intrinsic volumes, including surface area, integrated mean curvaure, and Euler characteristic. Often, the only information about the object is a digital image, which makes exact computations impossible. Moreover, most known estimation procedures are either biased or very complex. In this talk we shall consider a class of digital algorithms for estimation of all Minkowski tensors based on the Voronoi decomposition associated with the image. These algorithms can be shown to converge when the resolution goes to infinity. 

This is joint work with Markus Kiderlen and Daniel Hug.