Abstract

Forensic, security, and health-care processes defer to human observers when identifying visual differences beyond a certain degree of subtlety. For identikit, or passport-like, controls this typically involves extensive scrutiny of face-pair images. We tested forty-nine naive subjects to examine how they would mark differences seen between face pairs. In each trial, one pair of face images was displayed, side-by-side, for 20 s while the subjects marked where they saw differences by clicking the mouse within the right-hand image. The face pairs were specially selected from the BANCA image archive to best represent a wide variety of degrees of difference. Here, we report only trials involving two face pairs: a very difficult, and a very easy to distinguish face pair. From the result we are, nevertheless, able to classify participants into distinct sets of subject types by a parametric analysis of the temporal and spatial characteristics of each individual's click-trail. A graphical presentation of the analysis and classification is available. Although provisional, this classification might prove useful for assessing candidate scrutinisers for real-world tasks.

I documenti depositati in UnissResearch sono protetti dalle leggi che regolano il diritto d'autore