Darlow Luke Nicholas, Connan James
Appl Opt. 2015 Nov 1;54(31):9258-68. doi: 10.1364/AO.54.009258.
Optical coherence tomography provides a 3D representation of fingertip skin where surface and internal fingerprints are found. These fingerprints are topographically identical. However, the surface skin is prone to damage, distortion, and spoofing; and the internal fingerprint is difficult to access and extract. This research presents a novel scaling-resolution approach to fingerprint zone detection and extraction. Furthermore, a local-quality-based blending procedure is also proposed. The accuracy of the zone-detection algorithm is comparable to an earlier work, yielding a mean-squared error of 25.9 and structural similarity of 95.8% (compared to a ground-truth estimate). Blending the surface and internal fingerprints improved the National Institute of Science and Technology's Fingerprint Image Quality scores and the average maximum match scores (when matched against conventional surface counterparts). The fingerprint blending procedure was able to combine high-quality regions from both fingerprints, thus mitigating surface wrinkles and anomalous poor-quality regions. Furthermore, spoof detection via a surface-to-internal fingerprint comparison was proposed and tested.
光学相干断层扫描提供了指尖皮肤的三维表示，其中可以找到表面指纹和内部指纹。这些指纹在地形上是相同的。然而，表面皮肤容易受到损伤、变形和伪造；而内部指纹难以获取和提取。本研究提出了一种新颖的缩放分辨率方法用于指纹区域检测和提取。此外，还提出了一种基于局部质量的融合过程。区域检测算法的准确性与早期工作相当，均方误差为25.9，结构相似度为95.8%（与真实估计值相比）。将表面指纹和内部指纹进行融合提高了美国国家标准与技术研究院的指纹图像质量分数和平均最大匹配分数（与传统表面指纹匹配时）。指纹融合过程能够将来自两个指纹的高质量区域组合起来，从而减轻表面皱纹和异常低质量区域的影响。此外，还提出并测试了通过表面指纹与内部指纹比较进行伪造检测的方法。
