Triangulating Quadrilaterals on the Sphere: Application to Shape Analysis

Shape analysis aims at analysing the geometrical properties of a given object (e.g. organ or bone). Given a set of shapes, statistics can be estimated and modelled from this training population. Such models are useful to classify healthy and pathological subjects or to generate shape priors for image segmentation. Model construction involves establishment of corresponding points and shape parameterization is a common approach to solve this task, but inevitably introduces dis-tortions. To reduce these, we propose to triangulate the shapes’ parameterizations rather than the shapes themselves.