Enhancement of Fingerprint Images using Shape-Adapted Scale-Space Operators

Abstract

Automated Fingerprint Identification Systems (AFIS) rely on the ridge structure of a fingerprint for comparing it to other fingerprints in a database. The structure of current AFIS closely resembles the manual procedure used by human experts (McCabe et al., 1992), and can be briefly described as follows:

• Enhance the fingerprint image. This usually involves a filtering step to suppress noise. The output from this step can be either a better grey-level image or a binary image in which the ridge structure of the original is preserved.

• Extract the ridge structure of the fingerprint, i.e., build a graph-like representation of the skeletonized ridges and their incidence relations.

• Classify the fingerprint according to its ridge structure at a coarse scale. This reduces the search space, thus speeding up the procedure.

• Detect minutiae, i.e. singularities in the ridge structure, such as bifurcations and ridge endings. Then, build a feature vector of the fingerprint.

• Store the feature vectors in a database for later use and/or search for fingerprints with matching feature vectors (which will most likely correspond to the same person).