The Hong Kong Polytechnic University 3D Finger Knuckle Images Database

Traditional finger knuckle identification relies on the 2D imaging from finger knuckles. The usage of 3D finger knuckle information (surface normal vector, depth, or curvature) can enable more reliable characterization of finger knuckle patterns as the 3D information is expected to be invariant to the change of illuminations. In addition, the 3D imaging setups generally acquire 2D images and therefore simultaneous use of 3D and 2D finger knuckle images can be used to achieve a performance that may not be possible when either of these such information is employed alone. Printed photograph cannot reveal 3D information and therefore 3D finger knuckle imaging systems can easily detect such spoof attacks. It is extremely difficult to present replicas of 3D finger knuckle patterns as they require subjects to intentionally present their fingers under sophisticated imaging setups, unlike those required for 2D finger knuckle images that can also be acquired covertly.

Lack of any 3D finger knuckle images database in the literature has required us to acquire a new dataset. This database was acquired using the same imaging setup that we developed earlier for the contactless 3d fingerprint image acquisition. The final version 3D finger knuckle database has been acquired from 228 different subjects and among these 190 subjects have volunteered to provide second session's data. Each of these subjects provided six forefinger 3D knuckle images and six middle finger 3D knuckle images.

Sample Images

Raw images acquired from different subjects

Segmented images acquired from different subjects

3D reconstructed images

Download and Copyright

The database is made available freely for the researchers from Jan 2019 onwards. Interested researchers should follow following steps to acquire the 'The Hong Kong Polytechnic University 3D Finger Knuckle Images Database.