On the spatial inhomogeneity of fingerprint minutiae: A regression approach

Fingerprint minutiae distribution plays a critical role in studies such as fingerprint individuality for strengthening the scientific validity of fingerprint evidence and generating synthetic fingerprints for large-scale system evaluations. Fingerprint minutiae are not uniformly distributed as once assumed. Spatial inhomogeneity has been found in minutiae distribution, yet it is not clear what underlies the pattern and affects the random process. In this paper, we quantitatively study the spatial inhomogeneity of minutiae distribution with respect to the level-1 features of singular points and ridge orientation field. We propose a minutiae intensity function to measure the spatial inhomogeneity and a kernel regression model to characterize the relationship between minutiae intensity and the level-1 features. Our statistical experiments on benchmark fingerprint databases show that: 1) the examined level-1 features partially explain regular variation of minutiae spatial distribution; 2) there remain significant fraction of variations unexplained, suggesting additional propositions. These results provide better understanding of the experiential probability of a random correspondence for fingerprint minutiae distribution modelling.