A Locally Deformable Statistical Shape Model (LDSSM)

Abstract

In Chap. 2, we have introduced the basic concepts of statistical shape models, but we have also discussed their limits, especially that a large number of training shapes is needed in order to capture the full amount of intra-class shape variation. We propose a new way to address this limitation in Sects. 3.1 and 3.2 by introducing a Locally Deformable Statistical Shape Model (LDSSM) that makes better use of the available training data than a global model and thus greatly reduces the number of required training shapes. Furthermore, our approach has no need for any predefined segments and it does not introduce shape variations that cannot be explained through the training data. In Sect. 3.3, we will additionally integrate our LDSSM into an iterative framework that can be used to solve image segmentation problems.