Latent fingerprints raised from crime scenes often contain overlapping prints, which are difficult to separate and combine with the most advanced fingerprint combiners. Some methods have been proposed for separating overlapping fingerprints to allow fingerprint matchers to successfully match the fingerprints of the components. These methods are limited by the accuracy of the estimated targeting field, which is not reliable for poor quality overlapping latent fingerprints. In this work, we improve the robustness of the separation of fingerprints superimposed, particularly for low quality images. Our algorithm reconstructs the orientation fields of component prints by modeling the fingerprint orientation fields.
This additional marking is acceptable in forensics, where the first priority is to improve latent coincident accuracy. The efficacy of the proposed method has been evaluated not only on simulated superimposed impressions, but also on superimposed real latent fingerprint images. Compared with the available methods, the proposed algorithm is more effective in separating poor quality overlapping fingerprints and in improving overlapping fingerprint overlapping accuracy.
Fingerprints are widely used for personal authentication in both forensic and civilian applications. Given a fingerprint image, fingerprint pairs extract characteristics (for example, minutiae) from it and match the characteristics with the reference characteristic templates to identify or verify the identity associated with the fingerprint. Typically, the input image contains only one fingerprint. However, in practice, particularly in forensic medicine, two or more fingerprints may overlap one another, resulting in a superimposed fingerprint image.