Using Hair Follicles with Leg Geometry to Align Androgenic Hair Patterns

Identifying criminals and victims, including but not limited to masked gunmen, rioters, paedophiles, and victims of sexual offenses in digital images can be a challenging forensic task because often, neither their faces nor tattoos are observable. Though skin marks and vein patterns provide new ways to identify them, new biometric traits are still demanded because skin marks on some subjects are limited and clear vein patterns are difficult to be obtained from people with high concentration of body fat or melanin. They can be covered by androgenic hair. A recent paper indicated that androgenic hair patterns extracted from lower legs can be used for personal identification and demonstrated their effectiveness. However, it neglected the importance of alignment before matching two patterns. To alleviate this alignment problem, leg geometry was proposed for low resolution androgenic hair patterns. For aligning androgenic hair patterns in high resolution images which may be obtained from child sexual abuse images and rioter images taken by reporters, in this paper, a new feature, androgenic hair follicles, is proposed to combine with leg geometry to achieve high alignment accuracy. With this new feature, a new alignment algorithm is developed. Experiments on 1,138 high resolution images from 283 different legs show that the proposed alignment algorithm provides improvements of 5%-10% on different experimental settings.