Demonstrable and anatomy-driven knuckle identification via crease map segmentation

Vyas, R. and Williams, B. and Rahmani, H. and Boswell-Challand, R. and Jiang, Z. and Angelov, P. and Black, S. (2025) Demonstrable and anatomy-driven knuckle identification via crease map segmentation. Signal, Image and Video Processing, 19 (4): 318. ISSN 1863-1703

Text (RiteshPaper)
RiteshPaper.pdf - Accepted Version
Available under License Creative Commons Attribution.
Download (1MB)

Abstract

Images of the human hand can be effectively deployed to assist with the identification of the perpetrators of serious crimes. One of the prominent and distinguishing features of the human hand is found in the skin of the finger knuckle regions, which includes creases forming complex and distinctive patterns. Exploiting knuckle skin crease patterning in the identification of perpetrators requires manual labelling from expert anthropologists, which is both laborious and time-consuming. Existing approaches for automatic knuckle recognition work in a black-box manner without explicitly revealing the causes of a match or no match. Whereas, the court-room proceedings demand a more transparent and reproducible matching procedure driven from anatomy and comparison of skin creases. Hence, development of automated algorithms to segment (trace) the knuckle creases and compare them exclusively can make the whole process demonstrable and convincing. This paper proposes an effective framework for knuckle crease identification that can directly work on full hand dorsal images to (i) localize the knuckle regions effectively, (ii) segment (trace) the knuckle creases and (iii) effectively compare knuckles through the segmented crease maps. The novel matching of knuckle creases is achieved through explicit comparison of the creases themselves and is investigated with a large public dataset to demonstrate the potential of the proposed approach.