The impact of tattoo segmentation on the performance of tattoo matching

Tattoos are an important trait for criminal and victim identification and have been widely used by law enforcement agencies around the world. To facilitate it, tattoo images of prisoners, gangsters and other suspects are collected regularly. Up until now, these tattoo images are still manually annotated by law enforcement officers and text-based searching is used to retrieve tattoos of criminals and victims. Researchers have applied image-based matching methods to tattoos. Tattoo images collected from prisoners and crime scenes always include background and body sites other than tattoos. The aim of this paper is to evaluate how these non-tattoo regions impact on the performance of the image based matching methods. In this paper, SIFT and SURF are used as the matching methods. Four experimental settings are designed and over 4000 tattoo images are used in this evaluation. The experimental results pinpoint clearly that non-tattoo regions have negative impact on the matching performance and imply that tattoo matching will benefit from accurate tattoo detection.