An essential component of many medical image analysis protocols is the establishment and manipulation of feature correspondences. These image features can assume such forms spanning the range of functions of individual or regional pixel intensities to geometric structures extracted as a preprocessing segmentation step. Many algorithms focusing on the latter set of salient features attempt to reduce these structures to such geometric primitives as surfaces, curves and/or points for correspondence-based study. Although the latter geometric primitive forms the basis of many of these algorithms, unrealistic constraints such as assumptions of identical cardinality between point-sets hinder general usage. Furthermore, the local structure for certain point-sets derived from segmentation processes is often ignored. In this paper, we introduce a family of novel information-theoretic measures for pooint-set registration derived as a generalization of the well-known Shannon entropy known as the Havrda-Charvat-Tsallis entropy. This divergence measure permits a fine-tuning between robustness and sensitivity emphasis. In addition, we employ a directly manipulated free-form deformation (DMFFD) transformation model, a recently developed variant of the well-known FFD transformation model.