3D multi-object deformable templates based on moment invariants

We propose a new way of embedding shape distributions in a deformable template. These distributions rely on global shape descriptors corresponding to the 3D moment invariants. In opposition to usual Fourier-Like descriptors, moment invariants can be updated during deformations at a relatively low cost. The moment-based distributions are included in a framework allowing the management of several simultaneously deforming objects. This framework is dedicated to the segmentation of brain deep nuclei in 3D magnetic resonance images. Results are presented confirming the descriptors invariance relatively to translation, rotation and scale. Then, the slow variations of moment invariants in the shape space is studied using anatomical and synthetic objects. Most of them turn out to be especially stable relatively to inter-individual variability, which results in a high discrimination power for the moment-based distributions.