Application of Polar Harmonic Transforms to Fingerprint Classification

Accurate object classification is often confounded by the variation of the pose of the object. We show in this chapter how a set of recently introduced transforms, called the Polar Harmonic Transforms (PHTs), can be used to produce a set of features for rotation invariant fingerprint representation. For accurate classification, fingerprint images often need to be corrected for rotational differences. Determining an orientation reference for achieving this, however, often results in ambiguity and is hence not always reliable. PHTs allow rotation invariant representation of the fingerprint images and hence discard the need for reference detection. Experimental results indicate that the proposed PHT-based classification scheme yields results that are comparable with state-of-the-art methods.