Online variable topology type photovoltaic grid-connected inverter
Learning fingerprint reconstruction from minutiae to image
1. LEARNING FINGERPRINT RECONSTRUCTION: FROM MINUTIAE TO IMAGE
ABSTRACT
The set of minutia points is considered to be themost distinctive feature for fingerprint
representation and iswidely used in fingerprint matching. It was believed that the minutiae set
does not contain sufficient information to reconstructthe original fingerprint image from which
minutiae wereextracted. However, recent studies have shown that it is indeedpossible to
reconstruct fingerprint images from their minutiaerepresentations. Reconstruction techniques
demonstrate the needfor securing fingerprint templates, improving the template
interoperability,and improving fingerprint synthesis. But, there is stilla large gap between the
matching performance obtained fromoriginal fingerprint images and their corresponding
reconstructedfingerprint images. In this paper, the prior knowledge aboutfingerprint ridge
structures is encoded in terms of orientationpatch and continuous phase patch dictionaries to
improve thefingerprint reconstruction. The orientation patch dictionary isused to reconstruct the
orientation field from minutiae, while thecontinuous phase patch dictionary is used to reconstruct
the ridgepattern. Experimental results on three public domain databases(FVC2002 DB1_A,
FVC2002 DB2_A, and NIST SD4) demonstratethat the proposed reconstruction algorithm
outperformsthe state-of-the-art reconstruction algorithms in terms of both:1) spurious minutiae
and 2) matching performance with respectto type-I attack (matching the reconstructed fingerprint
againstthe same impression from which minutiae set was extracted) andtype-II attack (matching
the reconstructed fingerprint against adifferent impression of the same finger).