|
Abstract:
|
This talk presents the statistical analysis of elastic shape graphs. Each elastic shape graph is made up of nodes that are connected by several 2D or 3D curves or edges with arbitrary shapes. We develop a comprehensive Riemannian framework for computing shape metrics, shape geodesics, sample means, and tangent PCA for analyzing elastic graphs. One key challenge here is the registration of nodes across large networks and we develop novel multiscale representations of shape graphs to handle this challenge. Registration of nodes is performed and propagated from low to high-resolution representations, resulting in an efficient procedure for matching complex networks. These ideas are demonstrated using the STARE database of retinal images.
|
