Image-based approach to fingerprint acceptability assessment

Abstract

Reliable acceptability assessment of an image acquired by a fingerprint scanner is one of the major requirements in securing fingerprint-based authentication systems. As a realistic solution to this, we propose and discuss a quality assessment algorithm, which, without any additional sensing hardware, extracts and analyzes features observed in a single input image. The image acceptability judgment is made as a result of discriminant function analysis of the features obtained from (1) a spatial changing pattern of gray level, which reflects the difference in the characteristics of the substance that constitutes the object, and (2) the frequency pattern of the image, which shows the existence and density of "micro-structures" on the surface of the object, such as sweat pores. Experiments show that this software-based approach is quite effective in correctly assessing the attributes of the object on the scanner, while it is platform independent and sufficiently fast. © Springer-Verlag Berlin Heidelberg 2004.