Fingerprint explorer
Team SAFE
Fingerprint Explorer
How does a machine recognise a finger? Follow a fingerprint, step by step, all
the way to its constellation of minutiae.
A fingerprint is not compared like a photograph. To recognise it, software first reduces it to a few dozen characteristic points, the minutiae: places where a ridge stops (ridge ending) or splits in two (bifurcation). Their layout is enough to tell one finger from another.
This demo opens the black box. Choose two fingerprints, then follow the processing step by step, from the raw image to the minutiae: ridge contrast, region of interest, local orientation, Gabor filtering, binarisation, thinning to a skeleton… Each step is shown as an image and explained in a sentence.
Then comes the comparison. The program looks for minutiae whose layout, distances and orientations, is found again from one print to the other, and links them into a constellation. An animation rotates and slides the first constellation until it lies on top of the second. The verdict: the same finger, or not. Like any biometric system, it sometimes gets it wrong: see if you can catch it out.
The 800 fingerprints on offer are synthetic: none belongs to a real person, and all computation runs in your browser, with nothing sent to a server. An installation version, connected to a real sensor, lets visitors try it with their own fingers.
The synthetic fingerprints are generated using SFinGe (University of Bologna); part of the image processing is inspired by Raffaele Cappelli's notebook.
Contributors
Tanguy Gernot, Emmanuel Giguet