Abstract: Optical Coherence Tomography (OCT) is a non-invasive, high-resolution imaging technology that has recently been used in fingerprint acquisition. The captured external and internal fingerprints are robust to spoofing attacks and resistant to harsh skin conditions. The quality of the OCT fingerprints is limited by imaging mechanism and reconstruction. The paper proposes a fingerprint fusion network leveraging the concept that multiple fingerprints derived from OCT volume data can enhance one another. This network integrates cross-attention and quality optimization techniques to enhance the quality of OCT fingerprint images. Three fingerprints, two internal fingerprints and one external fingerprint reconstructed based on depth or intensity information, are utilized. The network introduces a cross-attention module and a quality index OCL to improve the feature representation capabilities and the retention of high-quality regions. Experimental results show that the fingerprint images obtained by the fusion method in this paper are superior in visual effects and quality scores. abstract environment.
Loading