Cross-matching contactless fingerprints against legacy contact-based fingerprints

1. A method for rectifying contactless fingerprint perspective distortion, comprising:

selecting an electronic, perspective distorted fingerprint sample;

generating an unwarped fingerprint sample by rectifying perspective distortions from the electronic, perspective distorted fingerprint sample by application of an unwarping transformation, where parameters of the unwarping transformation are determined by a deep convolutional neural network (DCNN) trained on a database comprising contactless fingerprint samples suffering from perspective distortions; and

generating a warp score map corresponding to the unwarped fingerprint sample, the warp score map indicating an amount of warp at each source location of the electronic, perspective distorted fingerprint sample.

2. The method of claim 1 , wherein the database comprises contact-based fingerprint samples suffering from geometric distortion, at least one contact-based fingerprint sample corresponding to individual ones of the contactless fingerprint samples, and wherein the DCNN is trained on the contactless fingerprint samples and the corresponding contact-based fingerprint samples.

3. The method of claim 1 , wherein the DCNN is trained to determine the parameters of the unwarping transform based upon a PCA-constrained model of warp.

4. The method of claim 1 , wherein the DCNN is trained to determine the parameters of the unwarping transform based upon a free grid model of warp.

5. The method of claim 1 , wherein the unwarping transformation is a thin plate spline (TPS) transformation.

6. The method of claim 1 , wherein one source of the perspective distortions results from a touchless camera photograph of a fingerprint.

7. A method for rectifying contactless fingerprint perspective distortion, comprising:

selecting an electronic, perspective distorted fingerprint sample;

generating an unwarped fingerprint sample by rectifying perspective distortions from the electronic, perspective distorted fingerprint sample by application of an unwarping transformation, where parameters of the unwarping transformation are determined by a deep convolutional neural network (DCNN) trained on a database comprising contactless fingerprint samples suffering from perspective distortions; and

generating an enhanced ridge map from the unwarped fingerprint sample, the enhanced ridge map generated utilizing a convolutional neural network (CNN) based ridge enhancer.

8. The method of claim 7 , comprising identifying a relationship between the electronic, perspective distorted fingerprint sample and a contact-based fingerprint based upon a comparison of the enhanced ridge map with the contact-based fingerprint.

9. The method of claim 7 , comprising determining a reconstruction loss associated with the enhanced ridge map based upon comparison of the enhanced ridge map and a ground truth ridge map associated with a contact-based fingerprint corresponding to the electronic, perspective distorted fingerprint sample.

10. The method of claim 9 , wherein the reconstruction loss is provided for supervised training of the DCNN for determining the parameters of the unwarping transformation.

11. The method of claim 10 , wherein the reconstruction loss is weighted based upon a warp score map corresponding to the unwarped fingerprint sample, the warp score map indicating an amount of warp at each source location of the electronic, perspective distorted fingerprint sample.

12. The method of claim 11 , wherein the reconstruction loss is weighted based upon a masked score map generated from the warp score map based upon a ground truth map of the corresponding contact-based fingerprint.

13. The method of claim 7 , wherein the DCNN is trained to determine the parameters of the unwarping transform based upon a free grid model of warp.

14. The method of claim 7 , wherein the unwarping transformation is a thin plate spline (TPS) transformation.

15. A system, comprising:

processing circuitry comprising a processor and memory; and

a fingerprint unwarping and ridge enhancement application executable by the processing circuitry, where execution of the fingerprint unwarping and ridge enhancement application causes the processing circuitry to:

identify warp parameters associated with a contactless fingerprint sample, the warp parameters estimated from the contactless fingerprint sample by a deep convolutional neural network (DCNN) trained on a database comprising contactless fingerprint samples suffering from perspective distortions; and

generate an unwarped fingerprint sample from the contactless fingerprint sample, the unwarped fingerprint sample generated using an unwarping transformation based upon the identified warp parameters; and

wherein a distortion rectification application causes the processing circuitry to generate an enhanced ridge map from the unwarped fingerprint sample utilizing a convolutional neural network (CNN) based ridge enhancer.

16. The system of claim 15 , wherein the warp parameters are PCA-constrained warp parameters (Θ PCA ) or free grid warp parameters (Θ FG ).

17. The system of claim 15 , wherein the unwarping transformation is a thin plate spline (TPS) transformation.

18. The system of claim 15 , wherein supervised training of the DCNN is provided based upon a reconstruction loss based upon comparison of the enhanced ridge map and a ground truth ridge map associated with a contact-based fingerprint corresponding to the contactless fingerprint sample.

19. The system of claim 18 , wherein the reconstruction loss is weighted based upon a warp score map corresponding to the unwarped fingerprint sample, the warp score map indicating an amount of warp at each source location of the contactless fingerprint sample.

20. The system of claim 15 , wherein the contactless fingerprint is obtained by the system via an imaging device.