Fingerprint Growth Prediction, Image Preprocessing and Multi-level Judgment Aggregation

Abstract

Finger growth is studied in the first part of the thesis and a method for growth prediction is presented. The effectiveness of the method is validated in several tests. Fingerprint image preprocessing is discussed in the second part and novel methods for orientation field estimation, ridge frequency estimation and image enhancement are proposed: the line sensor method for orientation estimation provides more robustness to noise than state of the art methods. Curved regions are proposed for improving the ridge frequency estimation and curved Gabor filters for image enhancement. The notion of multi-level judgment aggregation is introduced as a design principle for combining different methods at all levels of fingerprint image processing. Lastly, score revaluation is proposed for incorporating information obtained during preprocessing into the score, and thus amending the quality of the similarity measure at the final stage. A sample application combines all proposed methods of the second part and demonstrates the validity of the approach by achieving massive verification performance improvements in comparison to state of the art software on all available databases of the fingerprint verification competitions (FVC).