Nonparametric joint shape learning for customized shape modeling

We present a shape optimization approach to compute patient-specific models in customized prototyping applications. We design a coupled shape prior to model the transformation between a related pair of surfaces, using a nonparametric joint probability density estimation. The coupled shape prior forces with the help of application-specific data forces and smoothness forces drive a surface deformation towards a desired output surface. We demonstrate the usefulness of the method for generating customized shape models in applications of hearing aid design and pre-operative to intra-operative anatomic surface estimation.