Hierarchical fingerprint quality estimation scheme

The performance of an automatic fingerprint authentication system relies heavily on the quality of the captured fingerprint images. The error rates of such systems can significantly be decreased just by detecting and removing poor quality images right at the early stages. In this paper, an effective quality estimation scheme is proposed, which performs analysis in hierarchical fashion. This multilayered quality analysis comprises eight quality measures, three at global level (whole image)) and five at local level (block-wise), each one giving its respective quality score. The scores at their respective levels are fused into a single quality score by their corresponding fusion engines at two different levels. Fusion engines are based on the `Weighted Sum rule´. The scheme is also evaluated using `Sum rule´. The proposed scheme is experimented against fingerprint Verification Competition (FVC 2004) datasets. The experimental results show that this arrangement could correctly estimate and assign grades (good/acceptable/poor) to nearly 95% images in the dataset. Also, it is observed that this novel fusion-based hierarchical quality assessment model estimates fingerprint quality more effectively with `weighted sum rule´ than `sum rule´ and their individual counterparts.