Classical fingerprint analysts use binary images in the minutiae recognition and extraction processes. These obtained images do not have a sufficient quality that allows the extraction of the robust primitives, either during the contours detection or during skeletonization. In this study, a mathematical approach for the fingerprint curves modeling have been performed. The adopted smoothing technique is based on two geometric interpolation types adapted to fingerprint images. The obtained results reveal that the Bezier curve method has an error lower than that by the cubic spline method. The Bézier curve method has a RMS value of 0.035 pixels and an average maximum error of the order of 0.33 pixels. On the other hand, the cubic spline method has a RMS value about of 0.043 pixels and an average maximum error about of 0.37 pixels. We can see that the proposed method facilitates the design process of real-world objects and makes the fingerprint curves smooth.
Key words: Fingerprint image; Bézier curve method; Cubic spline method; Gabor filter; smooth curve.
|
