One method to enhance or analyze fingerprint images is to use a directionally applied enhancement operator. Critical to this operator is the correct local direction. Typically, the local direction is found by correlating lines of constant extent but different angles with the local gray level intensity. The extent of these directions is a trade- off between performance in high-curvature and noisy regions. If the extent is too short the correlation works well in high-curvature regions but fails in noisy ones. The longer extent does well in noisy regions but fails in high- curvature ones. This paper shows how the directional image enhancement can also be done in the frequency domain. This approach overcomes the noise and high-curvature limitation of the spatial domain methods.

Date Published: 28 January 1997
PDF: 11 pages
Proc. SPIE 2932, Human Detection and Positive Identification: Methods and Technologies, (28 January 1997); doi: 10.1117/12.265384

Published in SPIE Proceedings Vol. 2932:
Human Detection and Positive Identification: Methods and Technologies
Lisa A. Alyea; David E. Hoglund, Editor(s)