From Global to Local Statistical Shape Priors Novel Methods to Obtain Accurate Reconstruction Results with a Limited Amount of Training Shapes

This book proposes a new approach to handle the problem of limited training data. Common approaches to cope with this problem are to model the shape variability independently across predefined segments or to allow artificial shape variations that cannot be explained through the training data, both of which have their drawbacks. The approach presented uses a local shape prior in each element of the underlying data domain and couples all local shape priors via smoothness constraints. The book provides a sound mathematical foundation in order to embed this new shape prior formulation into the well-known variational image segmentation framework. The new segmentation approach so obtained allows accurate reconstruction of even complex object classes with only a few training shapes at hand.

Basics.- Statistical Shape Models (SSMs).- A Locally Deformable Statistical Shape Model (LDSSM).- Evaluation of the Locally Deformable Statistical Shape Model.- Global-To-Local Shape Priors for Variational Level Set Methods.- Evaluation of the Global-To-Local Variational Formulation.- Conclusion and Outlook.

Carsten Last received his diploma degree in computer and communications systems engineering (with distinction) from TU Braunschweig, Germany, in 2009. During his studies he worked as a student assistant in the area of speech enhancement at the Institute for Communications Technology at TU Braunschweig. From 2009 to 2015 he was a research assistant and PhD student at the Institute for Robotics and Process Control at TU Braunschweig, from which he received his doctorate degree in computer science in 2016 (summa cum laude). His research focused mainly on the areas of medical image processing and computer vision. Since 2015, he is working as a research engineer at Volkswagen AG in the area of autonomous driving.