Improving Sensor Interoperability between Contactless and Contact-Based Fingerprints Using Pose Correction and Unwarping

Figure 1

Preprocessing and postprocessing pipeline. The input is the grayscale image from the camera sensor. It is rotated and then processed by the object-detection network to find the fingertips. Those are cut out from the image and enhanced. Additionally, the fingertip’s sharpness is scored. All of this happens live as preprocessing. The results are saved and later used for the postprocessing. In postprocessing, the finger’s core is localized, and the finger contour is segmented. Those are used to correct the horizontal rotation of the finger, followed by a lateral rotation correction. Finally, the image is unwarped before the template comparison.