A Mathematical Toolkit to improve the Similarity Analysis for Fingerprint Images

While fingerprint biometrics are a mature technology, preprocessing steps such as image alignment and minutiae filtering still significantly impact their performance. To address these challenges, this paper introduces a comprehensive package of analytical methods designed to enhance fingerprint image similarity analysis. Initially, we present an alignment method based on partitioning and statistical image analysis, which does not rely on specific information, such as the location of singular points. Next, a minutiae filtering method is developed, focusing on detecting and removing edge minutiae, which tend to provide non-discriminative information. To establish an agnostic distance metric approach, we conducted extensive experiments on public fingerprint datasets, incorporating an ablation study to evaluate the impact of alignment and filtering methods. Our results demonstrate that the proposed combination of rotation and filtering significantly outperforms untreated images, providing an efficient and competitive solution that improves and contributes to the biometric analysis field.