Title: Comparison of face image quality metrics: Electronic and legacy mug shots
Authors: Kevin O’Connor, Gregory Hales, Jonathon Hight, Shimon Modi, Stephen Elliott
Publication date: 2011/4/11
Conference name: Computational Intelligence in Biometrics and Identity Management (CIBIM), 2011 IEEE Workshop on
Pages: 123-126
Publisher: IEEE
Description:
Abstract-Automated face recognition offers an effective method for identifying individuals. Face images have been used in a number of different applications, including driver’s licenses, passports and identification cards. To provide some form of standardization for photographs in these applications, ISO/IEC JTC 1 SC 37 have developed standardized data interchange formats to promote interoperability. There are many different publically available face databases available to the research community that are used to advance the field of …

G. Hales, Graduate Researcher & S. J. Elliott Ph.D.
In recent times it has become apparent that data sharing capabilities across state departments and law enforcement agencies is an issue, especially in terms of tracking, monitoring, and identifying persons of interest. There is a need to assess the image capture process, as well as sharing capabilities, and to incorporate commercially available facial recognition technology to reduce the errors in identifying persons of interest. The objective of this project is to evaluate legacy face images, assess and standardize the image capture process across Indiana Department of Corrections (DOC) agencies, integrate facial recognition to link face databases, and integrate mobile devices in law enforcement vehicles for face recognition. This research will lead to improvements in the efficiency and quality of the face image capture process in Indiana’s DOC facilities and BMV branches and facilitate image sharing capabilities across Indiana state agencies

The study examined the perception of individuals to their photograph, and whether their highest ranked self selected photograph had the highest image quality as determined by two face image quality algorithms. The research question was to determine if people, when given a choice, will select the best quality photograph to submit to the passport or consular (or other authorities). The study showed that people do not self-select their best quality images. There was no statistically significant relationship between an individual’s perception and a software based perception of image quality.

Testing and evaluation standards for the performance of face recognition algorithms do not sufficiently address the effect of the quality of the sensor or capture device. Standardized testing protocols do not incorporate sensor quality as one of the factors to be analyzed when benchmarking face recognition algorithms. The impact of camera sensor quality on FRS performance has not been examined extensively. In addition, no formal test design and protocol currently exists for examining this impact.

Previous research in face recognition has not addressed the effect image background color has on the accuracy and precision with which the FRS performs matches. The current data capture and storage standards for face recognition (INCITS 385) recommends a uniform 18% gray background to optimize FRS performance. But current identification schemes that use facial information (such as passports and drivers licenses) either require different color backgrounds or only specify background uniformity and not color.
The objectives of this research are two fold:

• Examine how using different quality sensors affect the capabilities of a face recognition algorithm.

• Examine impact of background colors on FRS performance

In recent times it has become apparent that data sharing capabilities across state departments and law enforcement agencies is an issue, especially in terms of tracking, monitoring, and identifying persons of interest. There is a need to assess the image capture process, as well as sharing capabilities, and to incorporate commercially available facial recognition technology to reduce the errors in identifying persons of interest. The BSPA Lab is currently working on a project in cooperation with the Indiana Department of Corrections (IDOC). The goal of this project is to integrate facial recognition technology into the IDOC system to allow for more efficient and accurate monitoring and identifying of fugitives and persons of interest.

There are currently a number of sub-projects going on in the lab that are necessary to determine if legacy and current mugshot images are NIST Mugshot Best Practices standard compliant and of good enough quality to be used in a facial recognition system. These projects will provide important feedback to the IDOC of the current state of their mugshot images as well as the image capture process so the IDOC can better understand if these legacy images will be usable.

The first project currently being conducted includes an image quality and standard compliance assessment of the electronic mugshot database, this database contains 49,694 mugshot images all of different individuals.

The second project will perform image quality as well as performance rates for 13,000 legacy mugshot images that were collected from the IDOC.

The third project will analyze the same 13,000 legacy mugshot images to determine percentages that already conform to the NIST standard, that can be optimized for standard compliance, and that cannot be optimized for standard compliance.

The last project involves assessing the data to determine image quality attributes that are the cause of most of the non-standard compliance, observe current image capture processes employed at IDOC facilities, and from the data and observation create a standard compliant layout and best practices guide to be implemented in IDOC facilities to capture standard compliant images in the future.

This paper outlines various research projects that have been conducted at Purdue University in the areas of environment, population, and devices. These areas are of interest as biometric technologies are currently being implemented in various business applications. The environmental research is concerned with the performance of a facial recognition algorithm at differing illumination levels. The second study looks at population, which examines differences in image quality with regard to population age.The third study outlines dynamic signature verification and the issues associated with signing on different digitizers.Download article (pdf) 

Title: Effects of illumination changes on the performance of Geometrix FaceVision® 3D FRS
Authors: Eric P Kukula, Stephen J Elliott, Roman Waupotitsch, Bastien Pesenti
Publication date: 2004/10/11
Conference name: Security technology, 2004. 38th annual 2004 international Carnahan conference on
Pages: 331-337
Publisher: IEEE
Description
Abstract This evaluation examined the effects of four frontal light intensities on the performance of a 3D face recognition algorithm, specifically testing the significance between an unchanging enrollment illumination condition (220-225 lux) and four different illumination levels for verification. The evaluation also analyzed the significance of external artifacts (ie glasses) and personal characteristics (ie facial hair) on the performance of the face recognition system (FRS). Collected variables from the volunteer crew included age,

Title: Evaluation of a facial recognition algorithm across three illumination conditions
Authors: Eric P Kukula, Stephen J Elliott
Publication date: 2004/9
Journal name: Aerospace and Electronic Systems Magazine, IEEE
Volume: 19
Issue: 9
Pages: 19-23
Publisher: IEEE
Description:
Abstract This work evaluated the performance of a commercially available face recognition algorithm for the verification of an individual’s identity pertaining to three enrollment illumination levels. Existing facial recognition technology from still or video sources is becoming a practical tool for law enforcement, security, and counter-terrorist applications despite the limitations of the current technology. At this time, facial recognition has been implemented in limited applications, but has not been exhaustively studied in adverse …

This study will evaluate a facial recognition algorithm in operational testing and evaluation at Purdue University Airport. This study will collect face images at the student flight operations center over a period of time to assess template aging, user habituation, as well as algorithm performance over time. The study will be
operational and unattended.

Title: Securing a restricted site-biometric authentication at entry point
Authors: Eric P Kukula, SJ Elliott
Publication date: 2003/10/14
Conference name: Security Technology, 2003. Proceedings. IEEE 37th Annual 2003 International Carnahan Conference on
Pages: 435-438
Publisher: IEEE
Description:
Abstract We evaluated the performance of a commercially available facial recognition algorithm for the verification of an individual’s identity (1: 1) across three illumination levels. Existing facial recognition technology from still or video sources is becoming a practical tool for law enforcement, security, and counter-terrorist applications despite the limitations of current technology. At this time, facial recognition holds promise and has been implemented
in limited applications, but has not been exhaustively researched in adverse conditions, …