Computer image retrieval by features: suspect identification

Abstract

Correct suspect identification of known offenders by witnesses deteriorates rapidly as more are examined in mugshot albums. Feature approaches, where mugshots are displayed in order of similarity to witnesses' descriptions, attempt to increase identification success by reducing this number. A methodology is proposed for system design and evaluation based on experiments, computer simulations, and four classes of system performance measures: identification performance, retrieval rank, tolerance performance, and feature quality. This was used to develop a system for 640 mugshots of known offenders. In three empirical tests, over 90% of witness searches resulted in suspects retrieved in the first eight mugshots.