Mixed media tattoo image matching using transformed edge alignment

摘要

Tattoos are a valuable biometric used by law enforcement to infer identity and affiliation of crime suspects and victims. Tattoo image databases are currently searched by matching text keywords to image meta-tags, which have limited utility to express the relevant content of tattoos. We propose a means to query tattoo image databases using a sketch drawing which may only abstractly resemble the tattoo. We use coherent point drift to align edges extracted from the sketch and tattoo images and score alignment quality by measuring similarity of shape context features. We demonstrate our approach for database image retrieval using the mixed media database from the NIST Tatt-C challenge and achieve superior retrieval performance compared to several other competitive edge matching approaches.