Kernel Density Estimation and Intrinsic Alignment for Knowledge-Driven Segmentation: Teaching Level Sets to Walk

We address the problem of image segmentation with statistical shape priors in the context of the level set framework. Our paper makes two contributions: Firstly, we propose a novel multi-modal statistical shape prior which allows to encode multiple fairly distinct training shapes. This prior is based on an extension of classical kernel density estimators to the level set domain. Secondly, we propose an intrinsic registration of the evolving level set function which induces an invariance of the proposed shape energy with respect to translation. We demonstrate the advantages of this multi-modal shape prior applied to the segmentation and tracking of a partially occluded walking person.