| Paper ID | B-3-3.4 |
| Paper Title |
Vein Pattern Visualisation using Conditional Generative Adversarial Networks |
| Authors |
Ali Keivanmarz, Hamid Sharifzadeh, Unitec Institute of Technology, New Zealand; Rachel Fleming, Institute of Environmental Science and Research (ESR), New Zealand |
| Session |
B-3-3: Recent Advances in Multimedia Security and Forensics |
| Time | Thursday, 10 December, 17:30 - 19:30 |
| Presentation Time: | Thursday, 10 December, 18:15 - 18:30 Check your Time Zone |
|
All times are in New Zealand Time (UTC +13) |
| Topic |
Multimedia Security and Forensics (MSF): |
| Abstract |
Utilising vein pattern as a biometric attribute for forensic identification in crime investigation has been challenging because vein patterns are almost undetectable in normal RGB images. Significant research efforts for uncovering vein patterns have been recently made based on various computational techniques such as artificial neural networks, optical vein disclosure, auto-encoders, etc. While some promising results have been achieved using these methods, comparing with the NIR reference images, these computational techniques are still struggling to provide reliable outcomes. In this paper, we propose a new method that performs vein pattern visualisation based on a conditional Generative Adversarial Network (GAN). GANs have shown very promising results on image translation tasks in other areas and therefore, for the first time, a specialised conditional GAN is proposed for translating colour RGB images to NIR images in this paper. The performance evaluation conducted on a small dataset shows the efficiency of our proposed method by uncovering over 80% of vein pixels in forearms of eleven subjects. |