On matching skulls to digital face images: A preliminary approach

Abstract

Forensic application of automatically matching skull with face images is an important research area linking biometrics with practical applications in forensics. It is an opportunity for biometrics and face recognition researchers to help the law enforcement and forensic experts in giving an identity to unidentified human skulls. It is an extremely challenging problem which is further exacerbated due to lack of any publicly available database related to this problem. This is the first research in this direction with a twofold contribution: (i) introducing the first of its kind skullface image pair database, IdentifyMe, and (ii) presenting a preliminary approach using the proposed semi-supervised formulation of transform learning. The experimental results and comparison with existing algorithms showcase the challenging nature of the problem. We assert that the availability of the database will inspire researchers to build sophisticated skull-to-face matching algorithms.

Learning A Shared Transform Model for Skull to Digital Face Image Matching

“…The two protocols are as follows: Figure 6: Bar graph representing the rank-5 identification accuracy for both the protocols on IdentifyMe dataset. SS-TL+HOG corresponds to the current state-of-the-art approach [8].…”

“…• Table 1 presents the rank-1 identification accuracy (%) of the proposed model on the IdentifyMe dataset for both the protocols, along with the comparative algo- Table 1: Rank-1 identification accuracy (%) on the Iden-tifyMe dataset. Baseline results have directly been taken from [8].…”

“…In order to address these limitations, this research proposes a novel Shared Transform Model (Figure 3), which learns a single "shared transform" for both skulls and digital face images, while reducing the intra-class variations. Experimental results on the publicly available skull dataset, Iden-tifyMe [8] demonstrate the efficacy of the proposed model for the task of skull to face matching. Figure 4 presents sample skull and digital face image pairs.…”

“…Baseline LBP [11] 28.5 14.2 DSIFT [2] 28.5 20.0 DL [7] 31.4 22.9 HOG [3] 34.2 25.7 US-TL + HOG [8] 37.1 34.2 SS-TL + HOG [8] 42.9 37.1 Proposed + HOG 51.4 42.9…”

“…Coupled with limited training samples, high intra-class variations render the problem of skull to digital face image matching challenging. Existing research in skull to photo matching has utilized transform learning [8]. The superior performance of transform learning models for other heterogeneous tasks with limited training samples [9] further make it an appropriate choice for addressing the given problem.…”