Fingerprint Alignment with Deformable Templates

The following paper introduces an extension to the deformable template paradigm and presents an application ofthe resulting image analysis method for fingerprint alignment. Instead of a scalar form of information that is to be stored at each of grid's nodes, we propose to use multidimensional vectors. To implement this idea, we formulate a novel rule for grid-object interactions that are necessary to execute grid-matching. The proposed concept enables an application of deformable templates to an analysis of vector fields, such as an orientation field, which can be derived for fingerprint images. We show that deformable templates with vector-valued nodes are a promising tool for compensation of nonlinear fingerprint deformations, so that they can appear useful in an image-alignment phase of fingerprint recognition procedures.