Kevin W. Bowyer - Publications

Copyright Notice.

This material is presented to ensure timely dissemination of scholarly and technical work. Copyright and all rights therein are retained by authors or by other copyright holders. All persons copying this information are expected to adhere to the terms and constraints invoked by each author's copyright. In most cases, these works may not be reposted without the explicit permission of the copyright holder.

All works below that appear in an IEEE publication have the copyright assigned to IEEE. Personal use of this material is permitted. However, permission to reprint/republish this material for advertising or promotional purposes or for creating new collective works for resale or redistribution to servers or lists, or to reuse any copyrighted component of this work in other works must be obtained from the IEEE. Similar conditions apply to works appearing in publications by other publishers.

PDF versions posted here may not have the final edits and page numbers of the version appearing in a journal or conference.

The research efforts reported on here have been supported all or in part by the NSF, CIA, DARPA, IARPA, Sandia National Labs and other agencies. Any opinions, findings, and conclusions or recommendations expressed are those of the author(s) and do not necessarily reflect the views of the sponsor(s).

Groups of papers by theme:

media forensics.

iris recognition.

face recognition.

studies involving identical twins.

studies involving human perception of image content.

publications with undergraduate co-authors.

data mining and classifier ensembles.

change detection in before / after aerial images.

ethical and social implications of technology.

ear biometrics.

medical image analysis.

edge detector evaluation.

object recognition based on functionality.

aspect graphs.

  • Analysis of diurnal changes in pupil dilation and eyelid aperture,
    Adam Czajka, Kevin W. Bowyer and Estefan Ortiz,
    IET Biometrics 7 (2), March 2018, 136-144.
    DOI link to this paper in IEEE Xplore.
    This work is inspired by the observation of surprising daily fluctuations in the number of valid iris code bits used to match irises in the NEXUS program operated by the Canadian Border Security Agency. These fluctuations have an impact on iris comparison scores but cannot be simply explained by pupil dilation, which does not have a clear pattern that would generalise to a population. To check if fluctuations in the number of valid iris code bits may be explained by eyelid aperture observed in a controlled, laboratory environment, the eyelid aperture was measured for 18 subjects participating in an acquisition every 2 h during the day. Simultaneously, the pupil dilation was measured to check the existence of a daily pattern for a population and for single subjects. There are two interesting outcomes of this work. First, there are statistically significant changes during the day in both pupil dilation and eyelid opening observed for individual subjects. Second, these changes do not generalise well into a common pattern for the group. Consequently, the diurnal fluctuations in the number of bits compared and the comparison score observed in the NEXUS program cannot be explained by changes in pupil dilation nor by eyelid aperture.

  • Recognition of image-orientation-based iris spoofing,
    Adam Czajka, Kevin W. Bowyer, Michael Krumdick, and Rosaura G. VidalMata,
    IEEE Transactions on Information Forensics and Security 12 (9), September 2017, 2184 - 2196.
    DOI link to this paper in IEEE Xplore.
    This paper presents a solution to automatically recognize the correct left/right and upright/upside-down orientation of iris images. This solution can be used to counter spoofing attacks directed to generate fake identities by rotating an iris image or the iris sensor during the acquisition. Two approaches are compared on the same data, using the same evaluation protocol: 1) feature engineering, using hand-crafted features classified by a support vector machine (SVM) and 2) feature learning, using data-driven features learned and classified by a convolutional neural network (CNN). A data set of 20 750 iris images, acquired for 103 subjects using four sensors, was used for development. An additional subject-disjoint data set of 1,939 images, from 32 additional subjects, was used for testing purposes. Both same-sensor and cross-sensor tests were carried out to investigate how the classification approaches generalize to unknown hardware. The SVM-based approach achieved an average correct classification rate above 95% (89%) for recognition of left/right (upright/upside-down) orientation when tested on subject-disjoint data and camera-disjoint data, and 99% (97%) if the images were acquired by the same sensor. The CNN-based approach performed better for same-sensor experiments, and presented slightly worse generalization capabilities to unknown sensors when compared with the SVM. We are not aware of any other papers on the automatic recognition of upright/upside-down orientation of iris images, or studying both hand-crafted and data-driven features in same-sensor and cross-sensor subject-disjoint experiments. The data sets used in this paper, along with random splits of the data used in cross-validation, are being made available.

  • Face Recognition Using Sparse Fingerprint Classification Algorithm,
    Tomas Larrain, John Bernhard, Domingo Mery and Kevin W. Bowyer,
    IEEE Transactions on Information Forensics and Security 12 (7), July 2017, 1646-1657.
    DOI link to this paper in IEEE Xplore.
    Unconstrained face recognition is still an open problem as the state-of-the-art algorithms have not yet reached high recognition performance in real-world environments. This paper addresses this problem by proposing a new approach called sparse fingerprint classification algorithm (SFCA). In the training phase, for each enrolled subject, a grid of patches is extracted from each subject's face images in order to construct representative dictionaries. In the testing phase, a grid is extracted from the query image and every patch is transformed into a binary sparse representation using the dictionary, creating a fingerprint of the face. The binary coefficients vote for their corresponding classes and the maximum-vote class decides the identity of the query image. Experiments were carried out on seven widely-used face databases. The results demonstrate that when the size of the data set is small or medium (e.g., the number of subjects is not greater than one hundred), SFCA is able to deal with a larger degree of variability in ambient lighting, pose, expression, occlusion, face size, and distance from the camera than other current state-of-the-art algorithms.

  • Lessons from Collecting a Million Biometric Samples,
    P. Jonathon Phillips, Patrick J. Flynn and Kevin W. Bowyer,
    Image and Vision Computing Journal 58, 96–107, February 2017.
    DOI link to this paper in ScienceDirect.
    Over the past decade, independent evaluations have become commonplace in many areas of experimental computer science, including face and gesture recognition. A key attribute of many successful independent evaluations is a curated data set. Desired aspects associated with these data sets include appropriateness to the experimental design, a corpus size large enough to allow statistically rigorous characterization of results, and the availability of comprehensive metadata that allow inferences to be made on various data set attributes. In this paper, we review a ten-year biometric sampling effort that enabled the creation of several key biometrics challenge problems. We summarize the design and execution of data collections, identify key challenges, and convey some lessons learned.

  • SREFI: Synthesis of Realistic Example Face Images,
    Sandipan Banerjee, John S. Bernhard, Walter J. Scheirer, Kevin W. Bowyer and Patrick J. Flynn,
    IEEE International Joint Conference on Biometrics, October 2017, Denver.
    DOI link to this paper in IEEE Xplore.
    In this paper, we propose a novel face synthesis approach that can generate an arbitrarily large number of synthetic images of both real and synthetic identities. Thus a face image dataset can be expanded in terms of the number of identities represented and the number of images per identity using this approach, without the identity-labeling and privacy complications that come from downloading images from the web. To measure the visual fidelity and uniqueness of the synthetic face images and identities, we conducted face matching experiments with both human participants and a CNN pre-trained on a dataset of 2.6M real face images. To evaluate the stability of these synthetic faces, we trained a CNN model with an augmented dataset containing close to 200,000 synthetic faces. We used a snapshot of this trained CNN to recognize extremely challenging frontal (real) face images. Experiments showed training with the augmented faces boosted the face recognition performance of the CNN.

  • Demography-based Facial Retouching Detection using Subclass Supervised Sparse Autoencoder,
    Aparna Bharati, Mayank Vatsa, Richa Singh and Kevin W. Bowyer, Xin Tong,
    IEEE International Joint Conference on Biometrics, October 2017, Denver.
    DOI link to this paper in IEEE Xplore.
    Digital retouching of face images is becoming more widespread due to the introduction of software packages that automate the task. Several researchers have introduced algorithms to detect whether a face image is original or retouched. However, previous work on this topic has not considered whether or how accuracy of retouching detection varies with the demography of face images. In this paper, we introduce a new Multi-Demographic Retouched Faces (MDRF) dataset, which contains images belonging to two genders, male and female, and three ethnicities, Indian, Chinese, and Caucasian. Further, retouched images are created using two different retouching software packages. The second major contribution of this research is a novel semi-supervised autoencoder incorporating "subclass" information to improve classification. The proposed approach outperforms existing state-of-the-art detection algorithms for the task of generalized retouching detection. Experiments conducted with multiple combinations of ethnicities show that accuracy of retouching detection can vary greatly based on the demographics of the training and testing images.

  • A Method for 3D Iris Reconstruction from Multiple 2D Near-Infrared Images,
    Diego Bastias; Claudio A. Perez; Daniel P. Benalcazar; Kevin W. Bowyer Claudio Perez, Diego Bastias and Kevin Bowyer,
    IEEE International Joint Conference on Biometrics, October 2017, Denver.
    DOI link to this paper in IEEE Xplore.
    The need to verify identity has become an everyday experience for most people. Biometrics is the principal means for reliable identification of people. Although iris recognition is the most reliable current technique for biometric identification, it has limitations because only segments of the iris are available due to occlusions from the eyelids, eyelashes, specular highlights, etc. The goal of this research is to study iris reconstruction from several 2D near infrared iris images, adding depth information to iris recognition. We expect that adding depth information from the iris surface will make it possible to identify people from a smaller segment of the iris. We designed a sensor for 2D near-infrared iris image acquisition. The method follows a pre-processing stage with the goal of performing iris enhancement, eliminating occlusions, reflections and extreme gray-level values, ending in iris texture equalization. The last step is the 3D iris model reconstruction based on several 2D iris images acquired at different angles. Results from each stage are presented.

  • LivDet Iris 2017 - Iris Liveness Detection Competition 2017,
    David Yambay, Benedict Becker, Naman Kohli, Daksha Yadav, Adam Czajka, Kevin W. Bowyer, Stephanie Schuckers, Richa Singh, Mayank Vatsa, Afzel Noore, D. Gragnaniello, C. Sansone, L. Verdoliva, Lingxiao He, Yiwei Ru, Haiqing Li, Nianfeng Liu, Zhenan Sun, Tieniu Tan,
    IEEE International Joint Conference on Biometrics, October 2017, Denver.
    DOI link to this paper in IEEE Xplore.
    Presentation attacks such as using a contact lens with a printed pattern or printouts of an iris can be utilized to bypass a biometric security system. The first international iris liveness competition was launched in 2013 in order to assess the performance of presentation attack detection (PAD) algorithms, with a second competition in 2015. This paper presents results of the third competition, LivDet-Iris 2017. Three software-based approaches to Presentation Attack Detection were submitted. Four datasets of live and spoof images were tested with an additional cross-sensor test. New datasets and novel situations of data have resulted in this competition being of a higher difficulty than previous competitions. Anonymous received the best results with a rate of rejected live samples of 3.36% and rate of accepted spoof samples of 14.71%. The results show that even with advances, printed iris attacks as well as patterned contacts lenses are still difficult for software-based systems to detect. Printed iris images were easier to be differentiated from live images in comparison to patterned contact lenses as was also seen in previous competitions.

  • Synthetic Minority Image Over-sampling Technique: How to Improve AUC for Glioblastoma Patient Survival Prediction,
    Renhao Liu, Lawrence Hall, Kevin Bowyer, Dmitry Goldgof, Robert Gatenby, Kaoutar Ben Ahmed,
    IEEE International Conference on Systems, Man, and Cybernetics, October 2017, Banf.
    DOI link to this paper in IEEE Xplore.
    Real-world datasets are often imbalanced, with an important class having many fewer examples than other classes. In medical data, normal examples typically greatly outnumber disease examples. A classifier learned from imbalanced data, will tend to be very good at the predicting examples in the larger (normal) class, yet the smaller (disease) class is typically of more interest. Imbalance is dealt with at the feature vector level (create synthetic feature vectors or discard some examples from the larger class) or by assigning differential costs to errors. Here, we introduce a novel method for over-sampling minority class examples at the image level, rather than the feature vector level. Our method was applied to the problem of Glioblastoma patient survival group prediction. Synthetic minority class examples were created by adding Gaussian noise to original medical images from the minority class. Uniform local binary patterns (LBP) histogram features were then extracted from the original and synthetic image examples with a random forests classifier. Experimental results show the new method (Image SMOTE) increased minority class predictive accuracy and also the AUC (area under the receiver operating characteristic curve), compared to using the imbalanced dataset directly or to creating synthetic feature vectors.

  • Spotting the Difference: Context Retrieval and Analysis for Improved Forgery Detection and Localization,
    Joel Brogan, Paolo Bestagini, Aparna Bharati, Allan Pinto, Daniel Moreira, Kevin Bowyer, Patrick Flynn, Anderson Rocha and Walter Scheirer,
    IEEE International Conference in Image Processing (ICIP), September 2017, China.
    arXiv link.
    As image tampering becomes ever more sophisticated and commonplace, the need for image forensics algorithms that can accurately and quickly detect forgeries grows. In this paper, we revisit the ideas of image querying and retrieval to provide clues to better localize forgeries. We propose a method to perform large-scale image forensics on the order of one million images using the help of an image search algorithm and database to gather contextual clues as to where tampering may have taken place. In this vein, we introduce five new strongly invariant image comparison methods and test their effectiveness under heavy noise, rotation, and color space changes. Lastly, we show the effectiveness of these methods compared to passive image forensics using Nimble [1], a new, state-of-the-art dataset from the National Institute of Standards and Technology (NIST).

  • Provenance Filtering for Multimedia Phylogeny,
    Allan Pinto, Daniel Moreira, Aparna Bharati, Joel Brogan, Kevin Bowyer, Patrick Flynn, Walter Scheirer and Anderson Rocha,
    IEEE International Conference in Image Processing (ICIP), September 2017, China.
    arXiv link.
    Departing from traditional digital forensics modeling, which seeks to analyze single objects in isolation, multimedia phylogeny analyzes the evolutionary processes that influence digital objects and collections over time. One of its integral pieces is provenance filtering, which consists of searching a potentially large pool of objects for the most related ones with respect to a given query, in terms of possible ancestors (donors or contributors) and descendants. In this paper, we propose a two-tiered provenance filtering approach to find all the potential images that might have contributed to the creation process of a given query q. In our solution, the first (coarse) tier aims to find the most likely "host" images --- the major donor or background --- contributing to a composite/doctored image. The search is then refined in the second tier, in which we search for more specific (potentially small) parts of the query that might have been extracted from other images and spliced into the query image. Experimental results with a dataset containing more than a million images show that the two-tiered solution underpinned by the context of the query is highly useful for solving this difficult task.

  • U-Phylogeny: Undirected Provenance Graph Construction In The Wild,
    Aparna Bharati, Daniel Moreira, Allan Pinto, Joel Brogan, Kevin Bowyer, Patrick Flynn, Walter Scheirer and Anderson Rocha,
    IEEE International Conference in Image Processing (ICIP), September 2017, China.
    arXiv link.
    Deriving relationships between images and tracing back their history of modifications are at the core of Multimedia Phylogeny solutions, which aim to combat misinformation through doctored visual media. Nonetheless, most recent image phylogeny solutions cannot properly address cases of forged composite images with multiple donors, an area known as multiple parenting phylogeny (MPP). This paper presents a preliminary undirected graph construction solution for MPP, without any strict assumptions. The algorithm is underpinned by robust image representative keypoints and different geometric consistency checks among matching regions in both images to provide regions of interest for direct comparison. The paper introduces a novel technique to geometrically filter the most promising matches as well as to aid in the shared region localization task. The strength of the approach is corroborated by experiments with real-world cases, with and without image distractors (unrelated cases).

  • Gender-From-Iris or Gender-From-Mascara?,
    Andrey Kuehlkamp, Benedict Becker and Kevin Bowyer,
    IEEE Winter Conference on Applications of Computer Vision (WACV), March 2017.
    DOI link to this paper in IEEE Xplore.
    Predicting a person's gender based on the iris texture has been explored by several researchers. This paper considers several dimensions of experimental work on this problem, including person-disjoint train and test, and the effect of cosmetics on eyelash occlusion and imperfect segmentation. We also consider the use of multi-layer perceptron and convolutional neural networks as classifiers, comparing the use of data-driven and hand-crafted features. Our results suggest that the gender-from-iris problem is more difficult than has so far been appreciated. Estimating accuracy using a mean of N person-disjoint train and test partitions, and considering the effect of makeup - a combination of experimental conditions not present in any previous work - we find a much weaker ability to predict gender-from-iris texture than has been suggested in previous work.

  • Handbook of Iris Recognition, second edition,
    Kevin W. Bowyer and Mark J. Burge, editors,
    Springer, 2016.
    Springer page.
    Jim Wayman's Afterword to the Second Edition.

  • Pitfalls In Studying "Big Data" From Operational Scenarios,
    Estefan Ortiz and Kevin W. Bowyer,
    IEEE 8th International Conference on Biometrics Theory, Applications and Systems (BTAS), September 2016.
    pdf of this paper.
    Analyzing a larger dataset is sometimes assumed, in itself, to give a greater degree of validity to the results of a study. In biometrics, analyzing an “operational” dataset is also sometimes assumed, in itself, to give a greater degree of validity. And so studying a large, operational biometric dataset may seem to guarantee valid results. However, a number of basic questions should be asked of any “found” big data, in order to avoid pitfalls of the data not being suitable for the desired analysis. We explore such issues using a large operational iris recognition dataset from the Canada Border Services Agency’s NEXUS program, similar to the dataset analyzed in the NIST IREX VI report.

  • Biometric Identification of Identical Twins: A Survey,
    Kevin W. Bowyer and Patrick J. Flynn,
    IEEE 8th International Conference on Biometrics Theory, Applications and Systems (BTAS), September 2016.
    pdf of this paper.
    The ability of biometric techniques to distinguish between identical twins is of interest for multiple reasons. The research literature touching on this topic is spread across a variety of areas. This survey pulls together the literature to date in this area, identifies available datasets for research, points out topics of uncertainty and suggests possible future research.

  • Template Aging in 3D and 2D Face Recognition,
    Ishan Manjani, Hakki Sumerkan, Patrick J. Flynn and Kevin W. Bowyer,
    IEEE 8th International Conference on Biometrics Theory, Applications and Systems (BTAS), September 2016.
    pdf of this paper.
    This is the first work to explore template aging in 3D face recognition. We use a dataset of images representing 16 subjects with 3D and 2D face images, and compare shortterm and long-term time-lapse matching accuracy. We find that an ensemble-of-regions approach to 3D face matching has much greater accuracy than whole-face 3D matching, or than a commercial 2D matcher. We observe a drop in accuracies with increased time lapse, most with whole-face 3D matching followed by 2D matching and the 3D ensemble of regions approach. Finally, we determine whether the difference in match quality arising with an increased time lapse is statistically significant.

  • On accuracy estimation and comparison of results in biometric research,
    Domingo Mery, Yuning Zhao and Kevin Bowyer,
    IEEE 8th International Conference on Biometrics Theory, Applications and Systems (BTAS), September 2016.
    pdf of this paper.
    The estimated accuracy of an algorithm is the most important element of the typical biometrics research publication. Comparisons between algorithms are commonly made based on estimated accuracies reported in different publications. However, even when the same dataset is used in two publications, there is a very low frequency of the publications using the same protocol for estimating algorithm accuracy. Using the example problems of face recognition, expression recognition and gender classification, we show that the variation in estimated performance on the same dataset across different protocols can be enormous. Based on these results, we make recommendations for how to obtain performance estimates that allow reliable comparison between algorithms.

  • Gender Classification from the Same Iris Code Used for Recognition,
    Juan Tapia, Claudio Perez and Kevin W. Bowyer,
    IEEE Transactions on Information Forensics and Security, 2016, to appear.
    pdf of this paper.
    ... This paper is the first to predict gender directly from the same binary iris code that could be used for recognition. We found that information for gender prediction is distributed across the iris, rather than localized in particular concentric bands. We also found that using selected features representing a subset of the iris region achieves better accuracy than using features representing the whole iris region. We used measures of mutual information to guide the selection of bits from the iris code to use as features in gender prediction. Using this approach, with a person-disjoint training and testing evaluation, we were able to achieve 89% correct gender prediction using the fusion of the best features of iris code from the left and the right eyes.

  • Detecting Facial Retouching Using Supervised Deep Learning,
    Aparna Bharati, Richa Singh, Mayank Vatsa and Kevin W. Bowyer,
    IEEE Transactions on Information Forensics and Security, September 2016, 1903-1913.
    pdf of this paper.
    Digitally altering, or "retouching", face images is a common practice for images on social media, photo sharing websites, and even identification cards when the standards are not strictly enforced. This research demonstrates the effect of digital alterations on the performance of automatic face recognition, and also introduces an algorithm to classify face images as original or retouched with good high performance. We first introduce two face image databases with unaltered and retouched images. Face recognition experiments performed on these databases show that when a retouched image is matched with its original image or an unaltered gallery image, the identification performance is considerably degraded, with a drop in matching accuracy of up to 25%. However, when images are retouched with the same style, the matching accuracy can be misleadingly high in comparison to matching original images. To detect retouching in face images, a novel supervised deep Boltzmann machine algorithm is proposed. It uses facial parts to learn discriminative features to classify face images as original or retouched. The proposed approach for classifying images as original or retouched yields an accuracy of over 87% on the datasets introduced in this paper and over 99% on three other makeup datasets used by previous researchers. This is a substantial increase in accuracy over the previous stateof- the-art algorithm [5] which has shown less than 50% accuracy in classifying original and retouched images from the ND-IIITD Retouched Faces database

  • An Analysis of 1-to-First Matching in Iris Recognition,
    Andrey Kuehlkamp and Kevin W. Bowyer,
    IEEE Workshop on Applications of Computer Vision, March 2016.
    pdf of this paper.
    ... A 1-to-First search terminates with the first below-threshold match that is found, whereas a 1-to-N search always finds the best match across all enrollments. We know of no previous studies that evaluate how the accuracy of 1-to-First search differs from that of 1-to-N search. Using a dataset of over 50,000 iris images from 2,800 different irises, we perform experiments to evaluate the relative accuracy of 1-to-First and 1-to-N search. We evaluate how the accuracy difference changes with larger numbers of enrolled irises, and with larger ranges of rotational difference allowed between iris images. We find that False Match error rate for 1-to- First is higher than for 1-to-N, and the the difference grows with larger number of enrolled irises and with larger range of rotation.

  • Critical Examination of the IREX VI Results,
    Kevin W. Bowyer and Estefan Ortiz,
    IET Biometrics 4 (4), 2015, 192-199.
    pdf of this paper.
    The authors analyse why Iris Exchange Report (IREX) VI conclusions about 'iris ageing' differ significantly from results of previous research on 'iris template ageing'. They observe that IREX VI uses a definition of 'iris ageing' that is restricted to a subset of International Organization for Standardization (ISO)-definition template ageing. They also explain how IREX VI commits various methodological errors in obtaining what it calls its 'best estimate of iris recognition ageing'. The OPS-XING dataset that IREX VI analyses for its 'best estimate of iris recognition ageing' contains no matches with Hamming distance >0.27. A 'truncated regression' technique should be used to analyse such a dataset, which IREX VI fails to do so, biasing its 'best estimate' to be lower-than-correct. IREX VI mixes Hamming distances from first, second and third attempts together in its regression, creating another source of bias towards a lower-than-correct value. In addition, the match scores in the OPS-XING dataset are generated from a '1-to-first' matching strategy, meaning that they contain a small but unknown number of impostor matches, constituting another source of bias towards an artificially low value for ageing. Finally, IREX VI makes its 'best estimate of iris recognition ageing' by interpreting its regression model without taking into account the correlation among independent variables. This is another source of bias towards an artificially low value for ageing. Importantly, the IREX VI report does not acknowledge the existence of any of these sources of bias. They conclude with suggestions for a revised, improved IREX VI.

  • Robust Detection of Textured Contact Lenses in Iris Recognition Using BSIF,
    James S. Doyle and Kevin W. Bowyer,
    IEEE Access, volume 3, 1672-1683, 14 September 2015, DOI: 10.1109/ACCESS.2015.2477470.
    link to this paper.
    This paper considers three issues that arise in creating an algorithm for the robust detection of textured contact lenses in iris recognition images. The first issue is whether the accurate segmentation of the iris region is required in order to achieve the accurate detection of textured contact lenses. Our experimental results suggest that accurate iris segmentation is not required. The second issue is whether an algorithm trained on the images acquired from one sensor will well generalize to the images acquired from a different sensor. Our results suggest that using a novel iris sensor can significantly degrade the correct classification rate of a detection algorithm trained with the images from a different sensor. The third issue is how well a detector generalizes to a brand of textured contact lenses, not seen in the training data. This paper shows that a novel textured lens type may have a significant impact on the performance of textured lens detection.

  • Dilation-Aware Enrollment for Iris Recognition,
    Estefan Ortiz, Kevin W. Bowyer and Patrick J. Flynn,
    IET Biometrics, DOI: 10.1049/iet-bmt.2015.0005. Available online: 13 October 2015.
    pdf of this paper.
    Iris recognition systems typically enroll a person based on a single "best" eye image. Research has shown that the probability of a false non-match result increases with increased difference in pupil dilation between the enrollment image and the probe image. Therefore, dilation-aware methods of enrollment should improve the accuracy of iris recognition. We examine a strategy to improve accuracy through a dilation-aware enrollment step that selects one or more enrollment images based on the observed distribution of dilation ratios for that eye. Additionally, we demonstrate that an image with median dilation is the optimal single eye image dilation-aware enrollment choice. Our results confirm that this dilation-aware enrollment strategy does improve matching accuracy compared to traditional single-image enrollment, and also compared to multi-image enrollment that does not take dilation into account.

  • Statistical Evaluation of Up-to-Three-Attempt Iris Recognition,
    Adam Czajka and Kevin W. Bowyer,
    IEEE International Conference on Biometrics Theory, Applications and Systems (BTAS 2015).
    pdf of this paper.
    Real-world biometric applications often operate in the context of an identity transaction that allows up to three attempts. That is, if a biometric sample is acquired and if it does not result in a match, the user is allowed to acquire a second sample, and if it again does not result in a match, the user is allowed to acquire a third sample. If the third sample does not result in a match, then the transaction is ended with no match. We report results of an experiment to determine whether or not successive attempts can be considered as independent samples from the same distribution, and whether and how the quality of a biometric sample changes in successive attempts. To our knowledge, this is the first published research to investigate the statistics of multi-attempt biometric transactions. We find that the common assumption that the attempt outcomes come from independent and identically distributed random variables in multi-attempt biometric transactions is incorrect.

  • Near-IR to Visible Light Face Matching: Effectiveness of Pre-Processing Options for Commercial Matchers,
    John Bernhard, Jeremiah Barr, Kevin W. Bowyer and Patrick J. Flynn
    IEEE International Conference on Biometrics Theory, Applications and Systems (BTAS 2015).
    pdf of this paper.
    ... Image pre-processing techniques can potentially be used to help reduce the differences between near-IR and visible light images, with the goal of improving matching accuracy. We evaluate the use of several such techniques in combination with commercial matchers and show that simply extracting the red plane results in a comparable improvement in accuracy. In addition, we show that many of the pre-processing techniques hinder the ability of existing commercial matchers to extract templates. We also make available a new dataset called Near Infrared Visible Light Database (ND-NIVL) consisting of visible light and near- IR face images with accompanying baseline performance for several commercial matchers.

  • Exploratory Analysis of an Operational Iris Recognition Dataset from a CBSA Border-Crossing Application,
    Estefan Ortiz and Kevin W. Bowyer,
    CVPR Biometrics Workshop, June 2015.
    pdf of this paper.
    This paper presents an exploratory analysis of an iris recognition dataset from the NEXUS border-crossing program run by the Canadian Border Services Agency. The distribution of the normalized Hamming distance for successful border-crossing transactions is examined in the context of various properties of the operational scenario. The effects of properties such as match score censoring and truncation, same-sensor and cross-sensor matching, sequence-dependent matching, and multiple-kiosk matching are illustrated. Implications of these properties of the operational dataset for the study of iris template aging are discussed.

  • Location Matters: A Study of the Effects of Environment on Facial Recognition for Biometric Security,
    Amanda Sgroi, Patrick Flynn, Kevin W. Bowyer and Hannah Garvey,
    Biometrics in the Wild Workshop 2015 (BWild 2015), Ljubljana, Slovenia, May 2015.
    pdf of this paper.
    The term “in the wild” has become wildly popular in face recognition research. The term refers generally to use of datasets that are somehow less controlled or more realistic. In this work, we consider how face recognition accuracy varies according to the composition of the dataset on which the decision threshold is learned and the dataset on which performance is then measured. We identify different acquisition locations in the FRVT 2006 dataset, examine face recognition accuracy for within-environment image matching and cross-environment image matching, and suggest a way to improve biometric systems that encounter images taken in multiple locations. We find that false non-matches are more likely to occur when the gallery and probe images are acquired in different locations, and that false matches are more likely when the gallery and probe images were acquired in the same location. These results show that measurements of face recognition accuracy are dependent on environment.

  • Trial Somaliland Voting Register De-duplication Using Iris Recognition,
    Kevin W. Bowyer, Estefan Ortiz and Amanda Sgroi,
    Biometrics in the Wild Workshop 2015 (BWild 2015), Ljubljana, Slovenia, May 2015.
    pdf of this paper.
    Face and fingerprint were used in de-duplication of the voter registration list for the 2010 Somaliland presidential election. Iris recognition was evaluated as a possible more powerful means of de-duplication of the voting register for the planned 2015 elections. On a trial dataset of 1,062 registration records, all instances of duplicate registration were detected and zero non-duplicates were falsely classified as duplicates, indicating the power of iris recognition for voting register deduplication. All but a tiny fraction of the cases were classified by automatic matching, and the remaining cases were classified by forensic iris matching. Images in this dataset reveal the existence of unusual eye conditions that consistently cause falsenon- match results. Examples are shown and discussed.

  • Lessons from collecting a million biometric samples,
    P. Jonathon Phillips, Patrick J. Flynn and Kevin W. Bowyer,
    IEEE International Conference on Automatic Face and Gesture Recognition (FG 2015), Ljubljana, Slovenia, May 2015.
    pdf of this paper.
    Over the past decade, independent evaluations have become commonplace in many areas of experimental computer science, including face and gesture recognition. A key attribute of many successful independent evaluations is a curated data set. Desired things associated with these data sets include appropriateness to the experimental design, a corpus size large enough to allow statistically rigorous characterization of results, and the availability of comprehensive metadata that allow inferences to be made on various data set attributes. In this paper, we review a ten-year biometric sampling effort that enabled the creation of several key biometrics challenge problems. We summarize the design and execution of data collections, identify key challenges, and convey some lessons learned.

  • Automatic facial attribute analysis via adaptive sparse representation of random patches,
    Domingo Mery and Kevin W. Bowyer,
    Pattern Recognition Letters 68, December 2015, 260-269.
    pdf of this paper.
    ... This paper addresses the problem of automated recognition of facial attributes by proposing a new general approach called Adaptive Sparse Representation of Random Patches (ASR+). ... Experiments were carried out on eight face databases in order to recognize facial expression, gender, race, disguise and beard. Results show that ASR+ deals well with unconstrained conditions, outperforming various representative methods in the literature in many complex scenarios.

  • Strong, Neutral or Weak: Exploring the Impostor Score Distribution,
    Amanda Sgroi, Patrick Flynn, Kevin Bowyer and P. Jonathon Phillips,
    IEEE Transactions on Information Forensics and Security, 10 (6), 1207-1220, June 2015.
    pdf of this paper.
    The strong, neutral, or weak (SNoW) face impostor pairs problem is intended to explore the causes and impact of impostor face pairs that are inherently strong (easily recognized as nonmatches) or weak (possible false matches). The SNoW technique develops three partitions within the impostor score distribution of a given data set. Results provide evidence that varying degrees of impostor scores impact the overall performance of a face recognition system. This paper extends our earlier work to incorporate improvements regarding outlier detection for partitioning, explores the SNoW concept for the additional modalities of fingerprint and iris, and presents methods for how to begin to reveal the causes of weak impostor pairs. We also show a clear operational difference between strong and weak comparisons as well as identify partition stability across multiple algorithms.

  • Face Recognition under Pose Variation with Local Gabor Features Enhanced by Active Shape and Statistical Models,
    Leonardo A. Cament, Francisco J. Galdames, Kevin W. Bowyer, and Claudio Perez,
    Pattern Recognition 48 (11), November 2015, 3371-3384.,
    pdf of this paper.
    Face recognition is one of the most active areas of research in computer vision. Gabor features have been used widely in face identification because of their good results and robustness. However, the results of face identification strongly depend on how different are the test and gallery images, as is the case in varying face pose. In this paper, a new Gabor-based method is proposed which modifies the grid from which the Gabor features are extracted using a mesh to model face deformations produced by varying pose. Also, a statistical model of the scores computed by using the Gabor features is used to improve recognition performance across pose. Our method incorporates blocks for illumination compensation by a Local Normalization method, and entropy weighted Gabor features to emphasize those features that improve proper identification. The method was tested on the FERET and CMU-PIE databases. Our literature review focused on articles with face identification with wide pose variation. Our results, compared to those of the literature review, achieved the highest classification accuracy on the FERET database with 2D face recognition methods. The performance obtained in the CMU-PIE database is among those obtained by the best methods published.

  • Statistical analysis of multiple presentation attempts in iris recognition,
    Adam Czajka and Kevin W. Bowyer,
    Second IEEE International Conference on Cybernetics (CYBCONF 2015), 24-26 June 2015.
    pdf of this paper.
    ... This paper presents experimental results showing uneven distributions of selected iris image quality metrics in the consecutive attempts in a biometric system that allows for multiple attempts to complete a transaction. ... One conclusion is that subjects rejected on the first attempt do, on average, improve the quality of their iris image on their second attempt. ... A second conclusion is that the probability of a subject being rejected on the second try is higher in average than the probability calculated for all subjects delivering their first samples. The latter finding contrasts with a common belief that each try in a single transaction can be assumed to be a draw from the same authentic distribution (and hence the rejection probabilities are equal in each try). ... To our knowledge, this paper presents the first research explaining the nature of multi-attempt iris recognition system and delivers conclusions that suggest that the default understanding of this process is too simplistic.

  • Framework for Active Clustering with Ensembles,
    Jeremiah R. Barr, Kevin W. Bowyer and Patrick J. Flynn,
    IEEE Transactions on Information Forensics and Security 9 (11), 1986-2001, November 2014.
    pdf of this paper.
    ... We introduce a novel semi-supervised framework for clustering face patterns into identity groups using minimal human interaction. This technique combines concepts from ensemble clustering and active learning to improve clustering accuracy. The framework actively queries the user for a soft link constraint between each pair of neighboring faces that are ambiguously matched according to the ensemble. We demonstrate the efficacy of our approach with the broadest evaluation of active face clustering algorithms to date. ...

  • Recognition of Facial Attributes Using Adaptive Sparse Representations of Random Patches,
    Domingo Mery and Kevin Bowyer,
    ECCV Workshop on Soft Biometrics, September 2014.
    pdf of this paper.
    ... This paper addresses the problem of automated recognition of facial attributes by proposing a new general approach called Adaptive Sparse Representation of Random Patches (ASR+). ...
    Best Paper Award.

  • Gender Classification from Iris Images Using Fusion of Uniform Local Binary Patterns,
    Juan E. Tapia, Claudio A. Perez and Kevin W. Bowyer,
    ECCV Workshop on Soft Biometrics, September 2014.
    pdf of this paper.
    ... Using a subject-disjoint test set, we are able to achieve over 91% correct gender prediction accuray using the texture of the iris. To our knowledge, this is the highest accuracy yet achieved for predicting gender from iris texture.

  • Cosmetic Contact Lenses and Iris Recognition Spoofing,
    Kevin W. Bowyer and James S. Doyle,
    IEEE Computer 47 (5), 96-98, May 2014.
    link to this paper in IEEE Xplore.
    Automatically detecting novel types of cosmetic contact lenses in iris images is a very difficult pattern-recognition problem, but experimental datasets that enable researchers to study this problem have recently become available.

  • Unraveling the Effect of Textured Contact Lenses on Iris Recognition,
    Daksha Yadav, Naman Kohli, James S. Doyle, Rich Singh, Mayank Vatsa and Kevin W. Bowyer,
    IEEE Transactions on Information Forensics and Security 9 (5), 851-862, May 2014.
    pdf of this paper.
    The presence of a contact lens, particularly a textured cosmetic lens, poses a challenge to iris recognition as it obfuscates the natural iris patterns. The main contribution of this paper is to present an in-depth analysis of the effect of contact lenses on iris recognition. Two databases, namely, IIITD Iris Contact Lens database and ND-Contact Lens database are prepared to analyze the variations caused due to contact lenses. We also present a novel lens detection algorithm that can be used to reduce the effect of contact lenses. The proposed approach outperforms other lens detection algorithms on the two databases and shows improved iris recognition performance.

  • Jim Thomas, Ahsan Kareem and Kevin W. Bowyer,
    Automated post-storm damage classification of low-rise building roofing systems using high resolution aerial imagery,
    IEEE Transactions on Geoscience and Remote Sensing 52 (7), 3851-3861, July 2014.
    link to this paper in IEEE Xplore.
    Techniques concerning postdisaster assessment from remotely sensed images have been studied by different research communities in the past decade. Such an assessment benefits a range of stakeholders, e.g., government organizations, insurance industry, local communities, and individual homeowners. This work explores detailed damage assessment on an individual building basis by utilizing supervised classification. In contrast with previous research efforts in the field, this work attempts at predicting the type of damages such as missing tiles, collapsed rooftop, and presence of holes, gaps, or cavities. ...

  • Jeffrey R. Paone, Patrick J. Flynn, P. Jonathon Philips, Kevin W. Bowyer, Richard W. Vorder Bruegge, Patrick J. Grother, George W. Quinn, Matthew T. Pruitt and Jason M. Grant,
    Double Trouble: Differentiating Identical Twins by Face Recognition,
    IEEE Transactions in Information Forensics and Security 9 (2), 285-295, February 2014.
    link to this paper in IEEE Xplore.
    Facial recognition algorithms should be able to operate even when similar-looking individuals are encountered, or even in the extreme case of identical twins. An experimental data set comprised of 17486 images from 126 pairs of identical twins (252 subjects) collected on the same day and 6864 images from 120 pairs of identical twins (240 subjects) with images taken a year later was used to measure the performance on seven different face recognition algorithms. Performance is reported for variations in illumination, expression, gender, and age for both the same day and cross-year image sets. Regardless of the conditions of image acquisition, distinguishing identical twins are significantly harder than distinguishing subjects who are not identical twins for all algorithms.

  • The Challenge of Face Recognition from Digital Point-and-Shoot Cameras,
    Ross Beveridge, Jonathon Phillips, David Bolme, Bruce Draper, Geof Givens, Yui Man Lui, Mohammad Nayeem Teli, Hao Zhang, W. Todd Scruggs, Kevin Bowyer, Patrick Flynn and Su Cheng,
    Biometrics Theory, Applications and Systems (BTAS) Sept 30 - Oct 2, 2013.
    pdf of this paper.
    Inexpensive point-and-shoot camera technology has combined with social network technology to give the general population a motivation to use face recognition technology. ... this paper introduces the Point-and-Shoot Face Recognition Challenge (PaSC). The challenge includes 9,376 still images of 293 people balanced with respect to distance to the camera, alternative sensors, frontal versus not-frontal views, and varying location. There are also 2,802 videos for 265 people: a subset of the 293. Verification results are presented for public baseline algorithms and a commercial algorithm for three cases: comparing still images to still images, videos to videos, and still images to videos.

  • SNoW: Understanding the Causes of Strong, Neutral and Weak Face Impostor Pairs,
    Amanda Sgroi, Kevin W. Bowyer, Patrick Flynn and P Jonathon Phillips,
    Biometrics Theory, Applications and Systems (BTAS) Sept 30 - Oct 2, 2013.
    pdf of this paper.
    The Strong, Neutral, or Weak Face Impostor Pairs problem was generated to explore the causes and impact of impostor face pairs that span varying strengths of scores. We develop three partitions within the impostor distribution for a given algorithm. The Strong partition contains image pairs that are easy to categorize as impostors. The Neutral partition contains image pairs that are less easily categorized as impostors. The Weak partition contains image pairs that are likely to cause false positives. ...

  • Similarity of Iris Texture Between Siblings,
    Will Vranderic and Kevin W. Bowyer,
    Biometrics Theory, Applications and Systems (BTAS) Sept 30 - Oct 2, 2013.
    pdf of this paper.
    Prior research has shown that twins' irises are similar enough in appearance for the untrained human observer to correctly determine whether an iris image pair comes from twins or from unrelated persons. We conducted a similar image pair classification study that asked participants to classify an image pair as ''siblings'' or ''unrelated''. We found that untrained human observers can classify pairs of siblings with over 57% accuracy using the appearance of the iris alone, without any proximal image content. This result is statistically greater than the accuracy of random guessing, which indicates that there is some degree of inherited texture similarity between siblings. This raises the question of how accurately one could classify siblings based on automated texture analysis.

  • Identity Verification Using Iris Images: Performance of Human Examiners,
    Kevin McGinn, Samuel Tarin and Kevin W. Bowyer,
    Biometrics Theory, Applications and Systems (BTAS) Sept 30 - Oct 2, 2013.
    pdf of this paper.
    We are not aware of any previous systematic investigation of how well human examiners perform at identity verification using the same type of images as acquired for automated iris recognition. This paper presents results of an experiment in which examiners consider a pair of iris images to decide if they are either (a) two images of the same eye of the same person, or (b) images of two different eyes, with the two different individuals having the same gender, ethnicity and approximate age. Results suggest that novice examiners can readily achieve accuracy exceeding 90% and can exceed 96% when they judge their decision as ''certain''. Results also suggest that examiners may be able to improve their accuracy with experience.

  • Effects of Iris Surface Curvature on Iris Recognition,
    Joseph Thompson, Patrick Flynn, Kevin W. Bowyer and Hector Santos-Villalobos,
    Biometrics Theory, Applications and Systems (BTAS) Sept 30 - Oct 2, 2013.
    pdf of this paper.
    To focus on objects at various distances, the lens of the eye must change shape to adjust its refractive power. This change in lens shape causes a change in the shape of the iris surface which can be measured by examining the curvature of the iris. This work isolates the variable of iris curvature in the recognition process and shows that differences in iris curvature degrade matching ability. To our knowledge, no other work has examined the effects of varying iris curvature on matching ability. ...

  • Variation in Accuracy of Textured Contact Lens Detection Based on Iris Sensor and Contact Lens Manufacturer,
    James S. Doyle, Kevin W. Bowyer and Patrick J. Flynn,
    Biometrics Theory, Applications and Systems (BTAS) Sept 30 - Oct 2, 2013.
    pdf of this paper.
    ... to date, the experimental results in this area have all been based on the same manufacturer of contact lenses being represented in both the training data and the test data and only one previous work has considered images from more than one iris sensor. Experimental results show that accuracy of textured lens detection can drop dramatically when tested on a manufacturer of lenses not seen in the training data, or when the iris sensor in use varies between the training and test data. These results suggest that the development of a fully general approach to textured lens detection is a problem that still requires attention.

  • A Linear Regression Analysis of the Effects of Age Related Pupil Dilation Change in Iris Biometrics,
    Estefan Ortiz, Kevin W. Bowyer, and Patrick J. Flynn
    Biometrics Theory, Applications and Systems (BTAS) Sept 30 - Oct 2, 2013.
    pdf of this paper.
    Medical studies have shown that average pupil size decreases linearly throughout adult life. Therefore, on average, the longer the time between acquisition of two images of the same iris, the larger the difference in dilation between the two images. Several studies have shown that increased difference in dilation causes an increase in the false nonmatch rate for iris recognition. Thus, increased difference in pupil dilation is one natural mechanism contributing to an iris template aging effect. We present an experimental analysis of the change in match scores in the context of dilation differences due to aging.

  • Template Aging Phenomenon In Iris Recognition,
    Samuel P. Fenker, Estefan Ortiz and Kevin W. Bowyer,
    IEEE Access 1, 266-274, May 16, 2013.
    link to open access version in IEEE Xplore.
    Biometric template aging is defined as an increase in recognition error rate with increased time since enrollment. It has been believed that template aging does not occur for iris recognition. However, several research groups have recently reported experimental results showing that iris template aging does occur. ...

  • Automated Classification of Contact Lens Type In Iris Images,
    James S. Doyle, Patrick J. Flynn and Kevin W. Bowyer,
    IAPR International Conference on Biometrics, June 2013.
    pdf of this paper.
    Textured cosmetic lenses have long been known to present a problem for iris recognition. It was once believed that clear, soft contact lenses did not impact iris recognition accuracy. However, it has recently been shown that persons wearing clear, soft contact lenses experience an increased false non-match rate relative to persons not wearing contact lenses. Iris recognition systems need the ability to automatically determine if a person is (a) wearing no contact lens, (b) wearing a clear prescription lens, or (c), wearing a textured cosmetic lens. This work presents results of the first attempt that we are aware of to solve this three-class classification problem. Results show that it is possible to identify with high accuracy (96.5%) the images in which a textured cosmetic contact lens is present, but that correctly distinguishing between no lenses and soft lenses is a challenging problem.

  • The Prediction of Young and Old Subjects From Iris Texture,
    Amanda Sgroi, Kevin W. Bowyer and Patrick J. Flynn,
    IAPR International Conference on Biometrics, June 2013.
    pdf of this paper.
    Researchers have previously studied the prediction of soft biometric attributes such as gender and ethnicity from iris texture images. We present the results of an initial study to predict the relative age of a person from such images. We conclude that is possible to categorize iris images as representing a young or older person at levels of accuracy statistically significantly greater than random chance. This suggests that there may in fact be age-related information available in the iris texture, and motivates further study of this topic.

  • The Impact of Diffuse Illumination on Iris Recognition,
    Amanda Sgroi, Kevin W. Bowyer and Patrick J. Flynn,
    IAPR International Conference on Biometrics, June 2013.
    pdf of this paper.
    Iris illumination typically causes specular highlighting both within the pupil and iris. This lighting variation is intended to be masked in the preprocessing stage. By removing or reducing these specular highlights, it is thought that a more accurate template could be made, improving the matching results. In an attempt to reduce these specular highlights we propose a diffuse illumination system. To determine if iris recognition performance is enhanced by this diffuse illumination system, we examine whether specular highlights were reduced within the pupil and iris, as well as analyze matching results obtained by several iris algorithms.

  • Detection of Contact-Lens-Based Iris Biometric Spoofs Using Stereo Imaging,
    Ken Hughes and Kevin W. Bowyer,
    Hawaii International Conference on System Sciences (HICSS 46), January 7-10, 2013.
    pdf of this paper.
    ... Thus the problem of determining if a person is wearing a cosmetic contact lens is transformed into the problem of classifying the estimated surface shape for the iris region. This is the first approach to analyze iris biometric images in the context of 3D shape.

  • Pose-Robust Recognition of Low-Resolution Face Images,
    Soma Biswas, Gaurav Aggarwal, Patrick J. Flynn and Kevin W. Bowyer,
    IEEE Transactions on Pattern Analysis and Machine Intelligence 35 (12), 3037-3049, December 2013.
    pdf of this paper.
    Face images captured by surveillance cameras usually have poor resolution in addition to uncontrolled poses and illumination conditions, all of which adversely affect performance of face matching algorithms. In this paper, we develop a completely automatic, novel approach for matching surveillance quality facial images to high resolution images in frontal pose which are often available during enrollment. The proposed approach uses multidimensional scaling to simultaneously transform the features from the poor quality probe images and the high quality gallery images in such a manner that the distances between them approximate the distances had the probe images been captured in the same conditions as the gallery images. Tensor analysis is used for facial landmark localization in the low-resolution uncontrolled probe images for computing the features. Thorough evaluation on the Multi-PIE dataset and comparisons with state-of-the-art super-resolution and classifierbased approaches are performed to illustrate the usefulness of the proposed approach. Experiments on surveillance imagery further signify the applicability of the framework. We also show the usefulness of the proposed approach for the application of tracking and recognition in surveillance videos.

  • Synthetic Eye Images for Pupil Dilation Mitigation,
    Robert Hasegawa, Estefan Ortiz, Kevin W. Bowyer, Louise Stark, Patrick J. Flynn and Ken Hughes,
    Biometrics Theory, Applications and Systems (BTAS) September 23-27, 2012.
    IEEE Xplore link for this paper.

  • Face Recognition from Video: A Review,
    Jeremiah Barr, Kevin W. Bowyer, Patrick Flynn, Soma Biswas,
    International Journal of Pattern Recognition and Artificial Intelligence 26 (5), August 2012.
    pdf of this paper.
    ... present a broad and deep review of recently proposed methods for overcoming the difficulties encountered in unconstrained settings. We also draw connections between the ways in which humans and current algorithms recognize faces. An overview of the most popular and difficult publicly available face video databases is provided to complement these discussions. Finally, we cover key research challenges and opportunities that lie ahead for the field as a whole.

  • Fast, Robust Feature-Based Matching for Automatic Image Registration in Disaster Response Applications,
    Jim Thomas, Ahsan Kareem and Kevin W. Bowyer,
    IEEE Interational Geoscience and Remote Sensing Symposium (IGARSS), July 2012.
    pdf of this paper.
    ... In this work, we propose a two-step approach to achieve fast and robust registration of before- / after-disaster aerial image pairs. This problem is difficult because there can be substantial change in the image content between the two images, and the time of day and lighting typically are different between the two images. ...

  • Analysis of Template Aging in Iris Biometrics,
    Samuel P. Fenker and Kevin W. Bowyer,
    IEEE Computer Society Biometrics Workshop, June 2012.
    pdf of this paper.
    It has been widely believed that biometric template aging does not occur for iris biometrics. We compare the match score distribution for short time-lapse iris image pairs, with a mean of approximately one month between the enrollment image and the verification image, to the match score distributions for image pairs with one, two and three years of time lapse. We find clear and consistent evidence of a template aging effect ...

  • Effects of Dominance and Laterality on Iris Recognition,
    Amanda Sgroi, Kevin W. Bowyer and Patrick Flynn,
    IEEE Computer Society Biometrics Workshop, June 2012.
    pdf of this paper.
    Eye dominance, the tendency to prefer to process visual input from one eye over the other, is a little discussed topic in iris biometrics. ... In this paper we explore the effects of eye dominance on iris recognition.

  • A Multi-Algorithm Analysis of Three Iris Biometric Sensors,
    Ryan Connaughton, Amanda Sgroi, Kevin W. Bowyer and Patrick J. Flynn,
    IEEE Transactions on Information Forensics and Security 7 (3), 919-931, June 2012.
    pdf of this paper.
    The issue of interoperability between iris sensors is an important topic in large-scale and long-term applications of iris biometric systems. This work compares three commercially available iris sensors and three iris matching systems and investigates the impact of cross-sensor matching on systems performance in comparison to single-sensor performance. Several factors which may impact single-sensor and cross-sensor perofrmance are analyzed ...

  • A Survey of Iris Biometrics Research: 2008-2010,
    Kevin W. Bowyer, Karen P. Hollingsworth and Patrick J. Flynn,
    in Handbook of Iris Recognition, Mark Burge and Kevin W. Bowyer, editors, Springer, 2013.
    pdf of this chapter.

  • Fusion of Face and Iris Biometrics,
    Ryan Connaughton, Kevin W. Bowyer and Patrick Flynn,
    in Handbook of Iris Recognition, Mark Burge and Kevin W. Bowyer, editors, Springer, 2013.
    pdf of this chapter.

  • Template Aging in Iris Biometrics: Evidence of Increased False Reject Rate in ICE 2006,
    Sarah Baker, Kevin W. Bowyer, Patrick J. Flynn and P. Jonathon Phillips,
    in Handbook of Iris Recognition, Mark Burge and Kevin W. Bowyer, editors, Springer, 2013.
    pdf of this chapter.
    license agreement for image dataset.

  • The Results of the NICE.II Iris Biometrics Competition,
    Kevin W. Bowyer,
    Pattern Recognition Letters 33 (8), 965-969, June 2012.
    pdf of this paper.
    ... NICE.II focused on performance in feature extraction and matching. The eight top-performing algorithms from NICE.II are considered, and suggestions are made for lessons that can be drawn from the results.

  • Human and Machine Performance on Periocular Biometrics Under Near-Infrared Light and Visible Light,
    Karen P. Hollingsworth, Shelby S. Darnell, Philip E. Miller, Damon L. Woodard, Kevin W. Bowyer and Patrick J. Flynn,
    IEEE Transactions on Information Forensics and Security 7 (2), 588-601, April 2012.
    pdf of this paper.
    ... Previous periocular research has used either visible light or near-infrared light images, but no prior research has directly compared the two illuminations using images with similar resolution. We conducted an experiment in which volunteers were asked to compare pairs of periocular images. ... We calculated performance of three computer algorithms on the periocular images. Performance for humans and computers was similar.

  • Best of Face and Gesture Recognition 2011,
    Rainer Stiefelhagen, Marian Stewart Bartlett and Kevin W. Bowyer,
    Image and Vision Computing Journal 30 (3), 135, March 2012.
    Link to this overview of the special issue.

  • A Sparse Representation Approach to Face Matching Across Plastic Surgery,
    Gaurav Aggarwal, Soma Biswas, Patrick J. Flynn and Kevin W. Bowyer,
    IEEE Workshop on Applications of Computer Vision, January 2012, Colorado Springs, CO.
    pdf of this paper.
    ... In this paper, we propose a novel approach to address the challenges involved in automatic matching of faces across plastic surgery variations. .... Extensive experiments conducted on a recently introduced plastic surgery database consisting of 900 subjects highlight the effectiveness of the proposed approach.

  • Predicting Good, Bad and Ugly Match Pairs,
    Gaurav Aggarwal, Soma Biswas, Patrick J. Flynn and Kevin W. Bowyer,
    IEEE Workshop on Applications of Computer Vision, January 2012, Colorado Springs, CO.
    pdf of this paper.
    ... The recently introduced GBU challenge problems indicates that even when the impact of most known factors is eliminated or significantly reduced by the data collection and experimental protocol, there can be significant variation in performance across different partitions of the data. ... In this paper, we investigate various image and facial characteristics that can account for the observed and significant difference in performance across these partitions. ...

  • Color Balancing for Change Detection in Multitemporal Images,
    Jim Thomas, Kevin W. Bowyer and Ahsan Kareem,
    IEEE Workshop on Applications of Computer Vision, January 2012, Colorado Springs, CO.
    pdf of this paper.
    narrated ppt on this work.
    ... In this paper we address color balancing for the purpose of change detection. ... We evaluated the proposed method against other state-of-the-art ones using a database consisting of aerial image pairs. The test image pairs were taken at different times, under different lighting conditions, and with different scene geometries and camera positions. On this database, our proposed approach outperformed other state-of-the-art algorithms.

  • Predicting Ethnicity and Gender from Iris Texture,
    Stephen Lagree and Kevin W. Bowyer,
    IEEE International Conference on Technologies for Homeland Security, November 2011, Boston, MA.
    pdf of this paper.
    Previous researchers have reported success in predicting ethnicity and predicting gender from features of the iris texture. This paper is the first to consider both problems from similar experimental approaches. Contributions of this work include greater accuracy than previous work on predicting ethnicity from iris texture, empirical evidence that suggests that gender prediction is harder than ethnicity prediction, and empirical evidence that ethnicity prediction is more difficult for females than for males.

  • Useful Features for Human Verification in Near-Infrared Periocular Images,
    Karen Hollingsworth, Kevin W. Bowyer and Patrick J. Flynn,
    Image and Vision Computing Journal 29 (11), October 2011, 707-715.
    pdf of this paper.
    ... We conducted two experiments to determine how humans analyze periocular images. In these experiments, we presented pairs of images and asked volunteers to determine whether the two images showed eyes from the same subject or from different subjects. ...
    Reprinted in the Journal of Intelligence Community Research and Development.

  • Multidimensional Scaling for Matching Low-resolution Face Images,
    Soma Biswas, Kevin W. Bowyer and Patrick J. Flynn,
    IEEE Transactions on Pattern Analysis and Machine Intelligence 34 (10), October 2012, 2019-2030.
    pdf of this paper.
    ... we propose a novel approach for matching low resolution probe images with higher resolution gallery images ... The proposed method simultaneously embeds the low resolution probe images and the high resolution gallery images in a common space such that the distances between them in the transformed space approximates the distances had both the images been of high resolution. The two mappings are learned simultaneously from high resolution training images using iterative majorization algorithm. Extensive evaluation of the proposed approach on the Multi-PIE dataset with probe image resolution as low as 8 x 6 pixels illustrates the usefulness of the method. We show that the proposed approach improves the matching performance significantly as compared to performing matching in the low-resolution domain or using super-resolution techniques to obtain a higher-resolution test image prior to recognition. ...

  • A Study of Face Recognition of Identical Twins By Humans,
    Soma Biswas, Kevin W. Bowyer and Patrick J. Flynn,
    International Workshop on Information Forensics and Security (WIFS 2011), December 2011 Foz do Iguacu, Brazil.
    pdf of this paper.
    In this work, we investigate human capability to distinguish between identical twins. If humans are able to distinguish between facial images of identical twins, it would suggest that humans are capable of identifying discriminating facial traits that can potentially be useful to develop algorithms for this very challenging problem. Experiments with different viewing times and imaging conditions are conducted to determine if humans viewing a pair of facial images can perceive if the image pairs belong to the same person or to a pair of identical twins. The experiments are conducted on 186 twin subjects, making it the largest such study in the literature to date.

  • Dilation Aware Multi-Image Enrollment for Iris Biometrics,
    Estefan Ortiz and Kevin W. Bowyer,
    International Joint Conference on Biometrics (IJCB 2011), October 2011.
    pdf of this paper.
    ... the probability of a false non-match increases when there is a considerable variation in pupil size between the enrolled eye image and the probe eye image. ... Our research examines a strategy to improve system performance by implementing a dilation-aware enrollment phase that chooses eye images based on their respective empirical dilation ratio distribution. ... Our results show that there is a noticeable improvement over the random scenario when pupil dilation is accounted for during the enrollment phase.

  • Twins 3D Face Recognition Challenge,
    Vipin Vijayan, Kevin W. Bowyer, Patrick Flynn, Di Huang, Liming Chen, Mark Hansen, Omar Ocegueda, Shishir Shah, Ioannis Kakadiaris,
    International Joint Conference on Biometrics (IJCB 2011), October 2011.
    pdf of this paper.
    Existing 3D face recognition algorithms have achieved high enough performances against public datasets like FRGC v2, that it is difficult to achieve further significant increases in recognition performance. However, the 3D TEC dataset is a more challenging dataset which consists of 3D scans of 107 pairs of twins that were acquired in a single session, with each subject having a scan of a neutral expression and a smiling expression. The combination of factors related to the facial similarity of identical twins and the variation in facial expression makes this a challenging dataset. We conduct experiments using state of the art face recognition algorithms and present the results. Our results indicate that 3D face recognition of identical twins in the presence of varying facial expressions is far from a solved problem, but that good performance is possible.

  • Predicting Ethnicity and Gender from Iris Texture
    (presentation only),
    Biometrics Consortium Conference (BCC), September 2011, Tampa, Florida.
    pdf of slides.

  • What Surprises Do Identical Twins Have for Identity Science?,
    Kevin W. Bowyer,
    IEEE Computer 44 (7), July 2011, 100-102.
    DOI link to this paper.
    Experiments with biometric datasets from identical twins are helping to shape future research in face and iris recognition.

  • Genetically Identical Irises Have Texture Similarity That Is Not Detected By Iris Biometrics,
    Karen Hollingsworth, Kevin W. Bowyer, Stephen Lagree, Samuel P. Fenker and Patrick J. Flynn,
    Computer Vision and Image Understanding 115, 1493-1502, 2011.
    pdf of this paper.
    As the standard iris biometric algorithm "sees" them, the left and right irises of the same person are as different as irises of unrelated people. Similarly, in terms of iris biometric matching, the eyes of identical twins are as different as irises of unrelated people. The left and right eyes of an individual or the eyes of identical twins are examples of genetically identical irises. In experiments with human observers viewing pairs of iris images acquired using an iris biometric system, we have found that there is recognizable similarity in the left and right irises of an individual and the irises of identical twins. This result suggests that iris texture analysis different from that performance in the standard iris biometric algorithm may be able to answer questions that iris biometrics cannot answer.

  • Improved Iris Recognition Through Fusion of Hamming Distance and Fragile Bit Distance,
    Karen Hollingsworth, Kevin W. Bowyer and Patrick J. Flynn,
    IEEE Transactions on Pattern Analysis and Machine Intelligence 33 (12), December 2011, 2465-2476.
    pdf of this paper.
    ... We find that the locations of fragile bits tend to be consistent across different iris codes of the same eye. We present a metric, called the fragile bit distance, which quantitatively measures the coincidence of the fragile bit patterns in two iris codes. We find that score fusion of fragile bit distances and Hamming distance works better for recognition than Hamming distance alone. To our knowledge, this is the first and only work to use the coincidence of fragile bit locations to improve the accuracy of matches.

  • A Cross-Sensor Evaluation of Three Commercial Iris Cameras for Iris Biometrics,
    Ryan Connaughton, Amanda Sgroi, Kevin W. Bowyer and Patrick J. Flynn,
    IEEE Computer Society Workshop on Biometrics, June 2011.
    pdf of this paper.
    ... This work presents experiments which compare three commercially available iris sensors and investigates the impact of cross-sensor matching on system performance. ...

  • Towards a robust automated hurricane damage assessment from high-resolution images,
    James Thomas, Ahsan Kareem, and Kevin W. Bowyer,
    13th International Conference on Wind Engineering (ICWE 13), July 2011.
    pdf of this paper.
    In the event of a natural disaster such as hurricane or earthquake, estimating the extent of damage is necessary for implementing fast and effective recovery measures. Images of affected areas are easily obtained through satellite or aerial sensors. The key objects of interest in such images are buildings, as damage directly impacts lives. This work aims to build a system capable of fine-grained damage analysis by comparing before and after storm images. ...

  • Distinguishing Identical Twins By Face Recognition,
    P. Jonathon Phillips, Patrick J. Flynn, Kevin W. Bowyer, Richard W. Vorder Bruegge, Patrick J. Grother, George W. Quinn, Matthew Pruitt,
    IEEE International Conference on Automatic Face and Gesture Recognition (FG 2011), March 2011, 185-192.
    pdf of this paper.
    This paper measures ability of face recognition algorithms to distinguish between identical twin siblings. The experimental dataset consists of images taken of 126 pairs of identical twins (252 people) collected on the same day and 24 pairs of identical twins (48 people) with images collected one year apart. Recognition experiments are conducted using three of the top submissions to the Multiple Biometric Evaluation (MBE) 2010 Still Face Track. ...

  • Experimental Evidence of a Template Aging Effect in Iris Biometrics,
    Sam Fenker and Kevin W. Bowyer,
    IEEE Computer Society Workshop on Applications of Computer Vision, January 2011.
    pdf of this paper.
    Baker et al recently presented the first published evidence of a template aging effect, using images acquired from 2004 through 2008 with an LG 2200 iris imaging system, representing a total of 13 subjects (26 irises). We report on a template aging study involving two different iris recognition algorithms, a larger number of subjects (43), a more modern imaging system (LG 4000), and over a shorter time-lapse (2 years). We also investigate the degree to which the template aging effect may be related to pupil dilation and/or contact lenses.
    license agreement for image dataset.

  • Detecting Questionable Observers Using Face Track Clustering,
    Jeremiah Barr, Kevin W. Bowyer and Patrick J. Flynn,
    IEEE Computer Society Workshop on Applications of Computer Vision, January 2011.
    pdf of this paper.
    We introduce the questionable observer detection problem: Given a collection of videos of crowds, determine which individuals appear unusually often across the set of videos. The algorithm proposed here detects these individuals by clustering sequences of face images. ...

  • Detecting and Ordering Salient Regions,
    Larry Shoemaker, Robert Banfield, Lawrence O. Hall, Kevin W. Bowyer and W. Philip Kegelmeyer,
    Data Mining and Knowledge Discovery 12 (1-2), January 2011, 259-290. http://dx.doi.org/10.1007/s10618-010-0194-6
    pdf of this paper.
    We describe an ensemble approach to learning salient regions from arbitrarily partitioned data. ... We combine a fast ensemble learning algorithm with scaled probabilistic majority voting in order to learn an accurate classifier ...

  • Human Perceptual Categorization of Iris Texture Patterns,
    Louise Stark, Kevin W. Bowyer and Stephen Siena,
    Fourth IEEE International Conference on Biometrics Theory, Applications and Systems (BTAS), September 2010. DOI: 10.1109/BTAS.2010.5634480
    pdf of this paper.
    We report on an experiment in which observers were asked to browse a set of 100 iris images and group them into categories based on similarity of overall texture appearance. Results indicate that there is a natural categorization of iris images into a small number of high-level categories, and then also into sub-categories. ...

  • Degradation of Iris Recognition Performance Due to Non-cosmetic Prescription Contact Lenses,
    Sarah Baker, Amanda Hentz, Kevin W. Bowyer and Patrick J. Flynn,
    Computer Vision and Image Understanding 114 (9), 1030-1044, September 2010.
    pdf of this paper.
    license agreement for image dataset.
    Many iris recognition systems operate under the assumption that non-cosmetic contact lenses have no or minimal effect on iris biometrics performance. ... This is the first study to document degraded iris biometrics performance with non-cosmetic contact lenses.

  • Human Versus Biometric Perception of Iris Texture,
    Presentation only of work done with Karen Hollingsworth, Steve Lagree, Sam Fenker and Patrick J. Flynn,
    Biometrics Consortium Conference (BCC), September 2010, Tampa, FL.
    pdf of slides.

  • Introduction to the Special Issue on Recent Advances In Biometrics,
    Kevin W. Bowyer,
    IEEE Transactions on Systems, Man and Cybernetics - Part A, 40 (3), May 2010, 434-436.
    pdf of this paper.
    As with BTAS 07, a biometrics-themed special issue of Systems Man and Cybernetics - Part A was organized following BTAS 08. However, unlike the SMC-A special section drawn from BTAS 07, this special issue had an "open" call for papers, meaning that submissions were not limited to papers presented at BTAS 08. ... A total of 31 submissions were received for the special issue. ... The ten papers that appear in this special issue represent the result of this process. Four of the ten papers appearing in this special issue are revised and extended versions of papers presented in the closing session of the BTAS 08 conference, a session which, by BTAS tradition, is reserved for the submissions that receive the overall best reviews from the conference program committee. The papers in this special issue cover a range of different biometric modalities, including face, ear, iris, signature and multi-modal. In the area of face recognition, there are papers dealing with 2D, 3D, and hand-drawn sketches. Also, the papers in this special issue range from relatively application oriented to relatively theoretical. One common theme is that each paper addresses an important current topic in biometrics research and makes a novel contribution to the state of the art.

  • Recent Advances in Biometric Systems: A Signal Processing Perspective,
    Natalia A. Schmid, Stephanie Schuckers, Jonathon Phillips and Kevin Bowyer,
    EURASIP Journal on Advances in Signal Processing, 2009. (Introduction to a special issue on Biometrics).
    (open access)
    We are pleased to receive a total of thirty-nine submissions to the special issue ... The final result is the set of fifteen papers that appear in this special issue. ...

  • Factors That Degrade the Match Distribution In Iris Biometrics,
    Kevin W. Bowyer, Sarah E. Baker, Amanda Hentz, Karen Hollingsworth, Tanya Peters and Patrick J. Flynn,
    Identity in the Information Society, 2 (3) 327-343, December 2009.
    DOI link. (open access)
    We consider three "accepted truths" about iris biometrics, involving pupil dilation, contact lenses and template aging. We also consider a relatively ignored issue that may arise in system interoperability. Experimental results from our laboratory demonstrate that the three accepted truths are not entirely true, and also that interoperability can involve subtle performance degradation. All four of these problems affect primarily the stability of the match, or authentic, distribution of template comparison scores rather than the non-match, or imposter, distribution of scores. In this sense, these results confirm the security of iris biometrics in an identity verification scenario. We consider how these problems affect the usability and security of iris biometrics in large-scale applications, and suggest possible remedies.

  • Iris Recognition Using Signal-level Fusion of Frames from Video,
    Karen P. Hollingsworth, Tanya Peters, Kevin W. Bowyer and Patrick J. Flynn,
    IEEE Transactions on Information Forensics and Security 4 (4), 837-848, December 2009.
    pdf of this paper.
    DOI link.
    We take advantage of the temporal continuity in an iris video to improve matching performance using signal-level fusion. From multiple frames of a frontal iris video, we create a single average image. ... No published prior work has shown any advantage of the use of video over still images in iris biometrics.

  • Using Fragile Bit Coincidence to Improve Iris Recognition,
    Karen P. Hollingsworth, Kevin W. Bowyer and Patrick J. Flynn,
    Third IEEE International Conference on Biometrics: Theory, Applications and Systems (BTAS 09), September 2009, Washington, DC.
    pdf of this paper.
    ... Previous research has shown that iris recognition performance can be improved by making these fragile bits. ... We find that the locations of fragile bits tend to be consistent across different iris codes of the same eye. We present a metrics, called the fragile bit distance, which quantitatively measures the coincidence of the fragile bit patterns in two iris codes. We find that score-fusion of fragile bit distance and Hamming distance works better for recognition than Hamming distance alone. This is the first and only work that we are aware of to use the coincidence of fragile bit locations to improve the accuracy of matches.

  • Stability of the Iris Match Distribution,
    Presentation only of work done with Karen Hollingsworth, Sarah Baker, Amanda Hentz, Tanya Peters and Patrick J. Flynn,
    Biometrics Consortium Conference (BCC), September 2009, Tampa, FL.
    pdf of slides.

  • The ND-IRIS-0405 Iris Image Dataset,
    Kevin W. Bowyer and Patrick J. Flynn,
    Notre Dame CVRL Technical Report.
    pdf of this report.
    The Computer Vision Research Lab at the University of Notre Dame began collecting iris images in the spring semester of 2004. The initial data collections used an LG 2200 iris imaging system for image acquisition. Image datasets acquired in 2004-2005 at Notre Dame with this LG 2200 have been used in the ICE 2005 and ICE 2006 iris biometric evaluations. The ICE 2005 iris image dataset has been distributed to over 100 research groups around the world. The purpose of this document is to describe the content of the ND-IRIS-0405 iris image dataset. This dataset is a superset of the iris image datasets used in ICE 2005 and ICE 2006.

  • FRVT 2006 and ICE 2006 Large-Scale Experimental Results,
    P. Jonathon Phillips, W. Todd Scruggs, Alice O'Toole, Patrick J. Flynn, Kevin W. Bowyer, Cathy L. Schott and Matthew Sharpe,
    IEEE Transactions on Pattern Analysis and Machine Intelligence 32 (5), May 2010, 831-846.
    pdf of this paper.
    DOI link.
    This paper describes the large-scale experimental results from the Face Recognition Vendor Test (FRVT) 2006 and the Iris Challenge Evaluation (ICE) 2006. ...

  • Efficacy of Damage Detection Measures from Aerial Images,
    Jim Thomas, Ahsan Kareem and Kevin W. Bowyer,
    11-th Americas Conference on Wind Engineering, June 2009.
    pdf of this paper.
    Estimating the extent of damage caused by natural disasters is necessary for implementing effective recovery measures. Aerial images of affected areas are easily obtained through satellite or aerial sensors. A careful analysis of images from before and after an event facilitates rapid detection and assessment of damage. Significant previous research has been done on developing measures to quantify the damage. In this study we evaluate the efficacy of existing change measures used to estimate damage. Determining the efficacy of these damage measures in definitive characterization of damage states is necessary for accurate automated assessment of windstorm damage.

  • Empirical Evidence for Correct Iris Match Score Degradation With Increased Time Lapse Between Gallery and Probe Images,
    Sarah Baker, Kevin W. Bowyer and Patrick J. Flynn,
    International Conference on Biometrics, June 2009, 1170-1179.
    pdf of this paper.
    We explore the effects of time lapse on iris biometrics using a data set of images with four years time lapse between the earliest and the most recent images of an iris (13 subjects, 26 irises, 1809 total images. We find that the average fractional distance for a match between two images of an iris taken four years apart is significantly larger than the match for images with only a few months time lapse between them. ... To our knowledge, this is the first and only experimental study of iris match scores under long (multi-year) time lapse.

  • Overview of the Multiple Biometric Grand Challenge,
    P. Jonathon Phillips, Patrick J. Flynn, J. Ross Beveridge, W. Todd Scruggs, Alice J. O'Toole, David Bolme, Kevin W. Bowyer, Bruce A. Draper, Geof. H. Givens, Yui Man Lui, Hassan Sahibzada, Joseph A. Scallan, and Samuel Weimer.
    International Conference on Biometrics, June 2009, 705-714.
    pdf of this paper.
    The goal of the Multiple Biometric Grand Challenge (MBGC) is to improve the performance of face and iris recognition technology from samples acquired under unconstrained conditions. The MBGC is organized into three challenge problems. Each challenge problem relaxes the constraints in different directions. ...

  • Image Averaging for Improved Iris Recognition,
    Karen Hollingsworth, Kevin W. Bowyer and Patrick J. Flynn,
    International Conference on Biometrics, June 2009, 1112-1121.
    pdf of this paper.
    We take advantage of the temporal continuity in an iris video to improve matching performance using signal-level fusion. From multiple frames of an iris video, we create a single average image. Our signal-level fusion method performs better that methods based on single still images, and better than previously published multi-gallery score-fusion methods. ...

  • Towards the Next Generation of Iris Biometrics Science,
    Kevin W. Bowyer, Patrick J. Flynn, Karen Hollingsworth, Sarah Baker and Sarah Ring,
    SPIE Newsroom, 2009. DOI link: http://dx.doi.org/10.1117/2.1200904.1595.
    pdf of this paper.
    (A SPIE Newsroom article that gives an overview of the paper and invited talk.)

  • Recent Research Results In Iris Biometrics,
    Karen Hollingsworth, Sarah Baker, Sarah Ring, Kevin W. Bowyer and Patrick J. Flynn,
    SPIE 7306B: Biometric Technology for Human Identification VI, April 2009.
    pdf of this paper.
    ... we have collected more than 100,000 iris images for use in iris biometrics research. Using this data, we have developed a number of techniques for improving recognition rates. These techniques include fragile bit masking, signal-level fusion of iris images, and detecting local distortions in iris texture. Additionally, we have shown that large degrees of dilation and long lapses of time between image acquisitions negatively impact performance.

  • Introduction to the Special Section of Best Papers from the 2007 Biometrics: Theory, Applications and Systems Conference,
    Kevin W. Bowyer,
    IEEE Transactions on Systems, Man and Cybernetics - Part A, 39 (1), January 2009, 2-3.
    pdf of this paper.
    DOI link.
    ... Over 100 papers were submitted to BTAS 07. ... The final result of this process is the set of five papers that appear in this special section. We are particularly fortunate in the way that the five papers in this special section illustrate the breadth of activities in current biometrics research. Face, fingerprint, iris, voice, handwriting, and multimodal biometrics are all represented. ...

  • Pupil Dilation Degrades Iris Biometric Performance,
    Karen Hollingsworth, Kevin W. Bowyer and Patrick J. Flynn,
    Computer Vision and Image Understanding, 113 (1), January 2009, 150-157.
    pdf of this paper.
    DOI link.
    ... We found that when the degree of dilation is similar at enrollment and recognition, comparisons involving highly dilated pupils result in worse recognition performance than comparisons involving constricted pupils. We also found that when the matched images have similarly highly dilated pupils, the mean Hamming distance of the match distribution increases and the mean Hamming distance of the non-match distribution decreases, bringing the distributions closer together from both directions. We further found that when matching enrollment and recognition images of the same person, larger differences in pupil dilation yield higher template dissimilarities, and so a greater chance of a false non-match. ...
    Our research group is part of the Multiple Biometric Grand Challenge and Iris Challenge Evaluation support teams.

  • The Best Bits In an Iris Code,
    Karen Hollingsworth, Kevin W. Bowyer and Patrick J. Flynn,
    IEEE Transactions on Pattern Analysis and Machine Intelligence 31 (6), 964-973, June 2009.
    pdf of this paper.
    ... The fractional Hamming distance weights all bits in an iris code equally. However, not all the bits in an iris code are equally useful. Our research is the first to present experiments documenting that some bits are more consistent than others. ... The inconsistencies are largely due to the coarse quantization of the phase response. Masking iris code bits corresponding to complex filter responses near the axes of the complex plane improves the separation between the match and nonmatch Hamming distance distributions.
    Our research group is part of the Multiple Biometric Grand Challenge and Iris Challenge Evaluation support teams.

  • Using Multi-Instance Enrollment to Improve Performance of 3D Face Recognition,
    Timothy C. Faltemier, Kevin W. Bowyer and Patrick J. Flynn,
    Computer Vision and Image Understanding 112 (2), November 2008, 114-125.
    Preprint pdf version of this paper.
    DOI link.
    This paper explores the use of multi-instance enrollment as a means to improve the performance of 3D face recognition. Experiments are performed using the ND-2006 3D face data set which contains 13,450 scans of 888 subjects. This is the largest 3D face data set currently available and contains a substantial amount of varied facial expression. Results indicate that the multi-instance enrollment outperforms a state-of-the-art component-based recognition approach ...

  • Detection of Iris Texture Distortions By Analyzing Iris Code Matching Results,
    Sarah Ring and Kevin W. Bowyer,
    Second IEEE International Conference on Biometrics: Theory, Applications and Systems (BTAS 08), September 2008, Washington, DC.
    pdf of this paper.
    Previous work in iris biometrics attempts to cope with occlusion by eyelids / eyelashes and with specular highlights through improved segmentation of the iris region. Our approach assumes that some local distortions of the iris texture are not detected at the segmentation stage, and that these generate corresponding regions of local distortion in the iris code derived from the image. We introduce an approach to detect such regions of local distortion in the iris code through analysis of the iris code matching results. We know of no previous work that attempts to detect distortions of iris texture through analyzing the iris code matching results.

  • Multi-factor Approach To Improving Recognition Performance In Surveillance-quality Video,
    Deborah Thomas, Kevin W. Bowyer and Patrick J. Flynn,
    Second IEEE International Conference on Biometrics: Theory, Applications and Systems (BTAS 08), September 2008, Washington, DC.
    DOI link.

  • Profile Face Detection: A Subset Multi-biometric Approach,
    James Gentile, Kevin W. Bowyer and Patrick J. Flynn,
    Second IEEE International Conference on Biometrics: Theory, Applications and Systems (BTAS 08), September 2008, Washington, DC.
    DOI link.

  • The Iris Challenge Evaluation 2005,
    P. Jonathon Phillips, Kevin W. Bowyer and Patrick J. Flynn, Xiaomei Liu and W. Todd Scruggs,
    Second IEEE International Conferenece on Biometrics: Theory, Applications and Systems (BTAS 08), September 2008, Washington, DC.
    DOI link.

  • A Region Ensemble for 3D Face Recognition,
    Timothy Faltemier, Kevin W. Bowyer and Patrick J. Flynn,
    IEEE Transactions on Information Forensics and Security, 3(1):62-73, March 2008.
    DOI link.
    ... we introduce a new system for 3D face recognition based on the fusion of results from a committee of regions that have been independently matched. ... Rank-one recognition rates of 97.2% and verification rates of 93.2% at 0.1% false accept rate are reported and compared to other methods published on the Face Recognition Grand Challenge v2 data set.

  • Image Understanding for Iris Biometrics: A Survey,
    Kevin W. Bowyer, Karen Hollingsworth and Patrick J. Flynn,
    Computer Vision and Image Understanding, 110(2), 281-307, May 2008.
    DOI link.
    ... Most research publications can be categorized as making their primary contribution to one of the four major modules in iris biometrics: image acquisition, iris segmentation, texture analysis and matching of texture representations. Other important research includes experimental evaluations, image databases, applications and systems, and medical conditions that may affect the iris. ...
    Our research group is part of the Multiple Biometric Grand Challenge and Iris Challenge Evaluation support teams.

  • Using Classifier Ensembles to Label Spatially Disjoint Data,
    Larry Shoemaker, Robert E. Banfield, Lawrence O. Hall, Kevin W. Bowyer and W. Philip Kegelmeyer,
    Information Fusion 9(1), 120-133, January 2008.
    pdf of this paper.
    We describe an ensemble approach to learning from arbitrarily partitioned data. ... We combine a fast ensemble learning algorithm with probabilistic majority voting in order to learn an accurate classifier from such data. ...

  • Learning to Predict Gender from Irises,
    Vince Thomas, Nitesh V. Chawla, Kevin W. Bowyer and Patrick J. Flynn,
    First IEEE International Conference on Biometrics: Theory, Applications, and Systems (BTAS 07), September 2007.
    pdf of this paper.
    This paper employs machine learning techniques to develop models that predict gender based on the iris texture features. ...

  • Guest Editorial: Introduction to the Special Issue on Recent Advances in Biometric Systems,
    Kevin W. Bowyer, Venu Govindaraju and Nalini Ratha,
    IEEE Transactions on Systems, Man and Cybernetics - B 37 (5), October 2007.
    pdf of this paper.
    We are pleased to present 14 papers in this special issue devoted to recent advances in biometric systems. ...

  • Comment on the CASIA version 1.0 Iris Dataset,
    P. Jonathon Phillips, Kevin W. Bowyer and Patrick J. Flynn,
    IEEE Transactions on Pattern Analysis and Machine Intelligence 29 (10), October 2007.
    pdf of this paper.
    We note that the images in the CASIA version 1.0 iris dataset have been edited so that the pupil area is replaced by a circular region of uniform intensity. We recommend that this dataset is no longer used in iris biometrics research ...

  • FRVT 2006 and ICE 2006 Large-Scale Results,
    P. J. Phillips, W. T. Scruggs, A. J. O'Toole, P. J. Flynn, K.W. Bowyer, C. L. Schott, and M. Sharpe.
    National Institute of Standards and Technology, NISTIR 7408, http://face.nist.gov, 2007.
    pdf of this report.
    This report describes the large-scale experimental results from the Face Recognition Vendor Test (FRVT) 2006 and the Iris Challenge Evaluation (ICE) 2006. ...
    Our research group is part of the Multiple Biometric Grand Challenge and Iris Challenge Evaluation support teams.

  • A Fast Algorithm for ICP-based 3D Shape Biometrics,
    Ping Yan and Kevin W. Bowyer,
    Computer Vision and Image Understanding, 107 (3), 195-202, September 2007.
    pdf of this paper.
    ... we present a novel approach, called "Pre-computed Voxel Nearest Neighbor," to reduce the computational time for shape matching in a biometrics context. The approach shifts the heavy computation burden to the enrollment stage, which is done offline. Experiments in 3D ear biometrics with 369 subjects and 3D face biometrics with 219 subjects demonstrate the effectiveness of our approach.

  • Biometric Recognition Using 3D Ear Shape,
    Ping Yan and Kevin W. Bowyer.
    IEEE Transactions on Pattern Analysis and Machine Intelligence 29 (8), 1297-1308, August 2007.
    pdf of this paper.
    ... We present a complete system for ear biometrics, including automated segmentation of the ear in a profile view image and 3D shape matching for recognition. We evaluated this system with the largest experimental study to date in ear biometrics, achieving a rank-one recognition rate of 97.8% for an identification scenario, and equal error rate of 1.2% for a verification scenario on a database of 415 subjects and 1,386 total probes.

  • Actively Exploring Face Space(s) for Improved Face Recognition,
    Nitesh V. Chawla and Kevin W. Bowyer,
    AAAI 2007, Vancouver, July 2007.
    pdf of this paper.
    We propose a learning framework that actively explores creation of face space(s) by selecting images that are complementary to the images already represented in the face space. We also construct ensembles of classifiers learned from such actively sampled image sets, which further provides improvement in the recognition rates. ...

  • Boosting Lite - Handling Larger Datasets and Slower Base Classifiers,
    Lawrence O. Hall, Robert E. Banfield, Kevin W. Bowyer and W. Philip Kegelmeyer,
    Multiple Classifier Systems (MCS) 2007, Prague, May 2007.
    pdf of this paper.
    ... we examine ensemble algorithms (Boosting Lite and Ivoting) that provide accuracy approximating a single classifier, but which require significantly fewer training examples. ...

  • A Comparison of Decision Tree Ensemble Creation Techniques,
    Robert E. Banfield, Lawrence O. Hall, Kevin W. Bowyer, and W. Philip Kegelmeyer,
    IEEE Transactions on Pattern Analysis and Machine Intelligence 29 (1), 173-180, January 2007.
    pdf of this paper. appendix to the paper.
    We experimentally evaluate bagging and seven other randomization-based approaches to creating an ensemble of decision tree classifiers. Statistical tests were performed on experimental results from 57 publicly available data sets. ...

  • Face Recognition Using 2D, 3D and Infra-Red: Is Multi-modal Better than Multi-sample?
    Kevin W. Bowyer, Kyong I. Chang, Patrick J. Flynn and Xin Chen,
    Proceedings of the IEEE, 94 (11), 2000-2012, November 2006.
    pdf of this paper.
    We compare the performance improvement obtained by combining three-dimensional or infra-red with normal intensity images (a multi-modal approach) to the performance improvement obtained by using multiple intensity images (a multi-sample approach). Combining results from different types of imagery gives significantly higher recognition rates than are obtained by using a single intensity image. However, significantly higher recognition rates are also obtained by combining results from multiple intensity images.

  • Multiple Nose Region Matching for 3D Face Recognition Under Varying Facial Expression,
    Kyong I. Chang, Kevin W. Bowyer, and Patrick J. Flynn,
    IEEE Transactions on Pattern Analysis and Machine Intelligence, 28 (10), 1695-1700, October 2006.
    pdf of this paper.
    An algorithm is proposed for 3D face recognition in the presence of varied facial expressions. It is based on combining the match scores from matching multiple overlapping regions around the nose. Experimental results are presented using the largest database employed to date in 3D face recognition studies, over 4,000 scans of 449 subjects. ...

  • Multi-modal Biometrics: An Overview,
    Kevin W. Bowyer, et al,
    Second Workshop on Multi-Modal User Authentication (MMUA 2006), May 2006, Toulouse, France.
    pdf of this paper.
    The topic of multi-modal biometrics has attracted strong interest in recent years. This paper categorizes approaches to multi-modal biometrics based on the biometric source, the type of sensing used, and the depth of collaborative interaction in the processing. This paper also attempts to identify some of the challenges and issues that confront research in multi-modal biometrics.
    This paper represents the invited talk given to open the first day of the workshop.

  • A Survey of Approaches and Challenges in 3D and Multi-modal 3D+2D Face Recognition,
    Kevin W. Bowyer, Kyong Chang, and Patrick J. Flynn,
    Computer Vision and Image Understanding 101 (1), January 2006, 1-15.
    pdf of this paper.
    ... This survey focuses on face recognition performed by matching models of the three-dimensional shape of the face, either alone or in combination with matching corresponding two-dimensional intensity images.
    This article was number one on the CVIU "Top 25" list for the quarters October - December 2005 and January - March of 2006, and in the top ten for seven consecutive quarters.

  • Experiments With an Improved Iris Segmentation Algorithm,
    Xiaomei Liu, Kevin W. Bowyer, and Patrick J. Flynn,
    Fourth IEEE Workshop on Automatic Identification Advanced Technologies (AutoID), October 2005, New York, 118-123.
    pdf of this paper.
    ... We have also developed and implemented an improved iris segmentation and eyelid detection stage of the algorithm, and experimentally verified the improvement in recognition performance using the collected dataset. Compared to Masek's original segmentation approach, our improved segmentation algorithm leads to an increase of over 6% in the rank-one recognition rate.

  • Eye Perturbation Approach for Robust Recognition of Inaccurately Aligned Faces,
    Jaesik Min, Kevin W. Bowyer, and Patrick J. Flynn,
    Fifth International Conference on Audio- and Video-Based Biometric Person Authentication (AVBPA 2005), July 2005, New York, 41-50.
    pdf of this paper.
    ... For improved performance and robustness to the eye location variation, we propose an eye perturbation approach that generates multiple face extractions from a query image by using the perturbed eye locations centered at the initial eye locations. ...

  • Infra-Red and Visible-Light Face Recognition,
    Xin Chen, Patrick J. Flynn, and Kevin W. Bowyer,
    Computer Vision and Image Understanding 99 (3), September 2005, 332-358.
    pdf of this paper.
    ... We find that in a scenario involving time lapse between gallery and probe, and relatively controlled lighting, (1) PCA-based recognition using visible images outperforms PCA-based recognition using infra-red images, and (2) the combination of PCA-based recognition using visible and infra-red imagery substantially outperforms either one individually...
    This article was number two on the CVIU "Top 25" list for July - September of 2005.

  • Ensembles of Classifiers from Spatially Disjoint Data,
    Robert E. Banfield, Lawrence O. Hall, Kevin W. Bowyer, and W. Philip Kegelmeyer,
    Springer-Verlag LNCS 3541: 6th International Workshop on Multiple Classifier Systems (MCS 2005), Monterey, CA, June 2005, 196-205.
    pdf of this paper.
    ... We describe an ensemble learning approach that accurately learns from data which has been partitioned according to the arbitrary spatial requirements of a large-scale simulation wherein classifiers may be trained only the data local to a given partition. As a result, the class statistics can vary from partition to partition; some classes may even be missing from some partitions.

  • Empirical Evaluation of Advanced Ear Biometrics,
    Ping Yan and Kevin W. Bowyer,
    Workshop on Empirical Evaluation Methods in Computer Vision, San Diego, CA, June 2005.
    pdf of this paper.
    We present results of the largest experimental investigation of ear biometrics to date. Approaches considered include a PCA ("eigen-ear") approach with 2D intensity images, achieving 63.8% rank-one recognition; a PCA approach with range images, achieving 55.3% Hausdorff matching of edge images from range images, achieving 67.5% and ICP matching of the 3D data, achieving 98.7%. ICP based matching not only achieves the best performance, but also shows good scalability with size of dataset. The data set used represents over 300 persons, each with images acquired on at least two different dates. In addition, the ICP-based approach is further applied on an expanded data set of 404 subjects, and achieves 97.5% rank one recognition rate. In order to test the robustness and variability of ear biometrics, ear symmetry is also investigated. In our experiments around 90% of peoples' right and left ears are symmetric.

  • Overview of the Face Recognition Grand Challenge,
    P. Jonathon Phillips, Patrick J. Flynn, Todd Scruggs, Kevin W. Bowyer, Jin Chang, Kevin Hoffman, Joe Marques, Jaesik Min, and William Worek,
    Computer Vision and Pattern Recognition (CVPR 2005), San Diego, June 2005, I:947-954.
    pdf of this paper.
    Our research group is part of the Face Recognition Grand Challenge support team.

  • Random Subspaces and Subsampling for 2-D Face Recognition,
    Nitesh V. Chawla and Kevin W. Bowyer,
    Computer Vision and Pattern Recognition (CVPR 2005) , San Diego, June 2005, II: 582-589.
    pdf of this paper.
    Random subspaces are a popular ensemble construction technique that improves the accuracy of weak classifiers. It has been shown, in different domains, that random subspaces combined with weak classifiers such as decision trees and nearest neighbor classifiers can provide an improvement in accuracy. In this paper, we apply the random subspace methodology to the 2-D face recognition task. The main goal of the paper is to see if the random subspace methodology can do as well, if not better, than the single classifier constructed on the tuned face space. We also propose the use of a validation set for tuning the face space, to avoid bias in the accuracy estimation. In addition, we also compare the random subspace methodology to an ensemble of subsamples of image data. This work shows that a random subspaces ensemble can outperform a well-tuned single classifier for a typical 2-D face recognition problem. The random subspaces approach has the added advantage of requiring less careful tweaking.

  • An Evaluation of Multi-modal 2D+3D Face Biometrics,
    Kyong I. Chang, Kevin W. Bowyer, and Patrick J. Flynn,
    IEEE Transactions on Pattern Analysis and Machine Intelligence 27 (4), April 2005, 619-624.
    pdf of this paper.
    We report on the largest experimental study to date in multi-modal 2D+3D face recognition ... Major conclusions are: (1) 2D and 3D have similar recognition performance when considered individually, (2) Combining 2D and 3D results using a simple weighting scheme outperforms either 2D or 3D alone, (3) Combining results from two or more 2D images using a similar weighting scheme also outperforms a single 2D image, and (4) Combined 2D+3D outperforms the multi-image 2D result. This is the first (so far, only) work to present such an experimental control to substantiate multi-modal performance improvement."

  • Ensemble Diversity Measures and Their Application to Thinning,
    Robert E. Banfield, Lawrence O. Hall, Kevin W. Bowyer, and W. Philip Kegelmeyer,
    Information Fusion 6 (1), March 2005, 49-62.
    pdf of this paper.
    ... We evaluate thinning algorithms on ensembles created by several techniques on 22 publicly available datasets. When compared to other methods, our percentage correct diversity measure algorithm shows a greater correlation between the increase in voted ensemble accuracy and the diversity value. ... Finally, the methods proposed for thinning again show that ensembles can be made smaller without loss in accuracy.

  • Improved Range Image Segmentation by Analyzing Surface Fit Patterns,
    Jaesik Min and Kevin W. Bowyer,
    Computer Vision and Image Understanding 97(2), February 2005, 242-258.
    pdf of this paper.
    We propose a new approach to range image segmentation of planar and curved surface scenes. Our method is mainly an extended design of an existing algorithm, which was guided by a framework of performance evaluation. We choose the range segmentation algorithm developed by Jiang and Bunke as our baseline algorithm, which is fast and has shown relatively high performance in several experimental performance evaluation studies. We analyze the types of errors made by the algorithm, propose design modifications to decrease the error rate, and experimentally verify that the new approach achieves statistically significant performance improvement. Whereas the baseline algorithm applies the edge-linking uniformly to all edge pixels to segment a region, the modified algorithm selects high potential edge areas in the region by analyzing the surface fit pattern and gives priority of edge-linking to those areas. The contributions of this work are (1) an improved algorithm for segmentation of range images of both planar and curved surface scenes, and (2) a demonstration of using empirical performance evaluation to guide algorithm design and modification to achieve better performance.

  • The Human ID Gait Challenge Problem: Data Sets, Performance, and Analysis,
    Sudeep Sarkar, P. Jonathon Phillips, Zongyi Liu, Isidro Robledo, Patrick Grother, and Kevin W. Bowyer,
    IEEE Transactions on Pattern Analysis and Machine Intelligence 27 (2), February 2005, 162-177.
    pdf of this paper.
    ... To provide a means for measuring progress and characterizing the properties of gait recognition, we introduce the HumanID Gait Challenge Problem. The challenge problem consists of a baseline algorithm, a set of 12 experiments, and a large data set. The baseline algorithm estimates silhouettes by background subtraction and performs recognition by temporal correlation of silhouettes. The 12 experiments are of increasing difficulty, as measured by the baseline algorithm, and examine the effects of five covariates on performance. ...

  • Comments on "A Parallel Mixture of SVMs for Very Large Scale Problems,"
    Xiaomei Liu, Lawrence O. Hall, and Kevin W. Bowyer,
    Neural Computation 16 (7), July 2004, 1345-1351.
    pdf of this paper.
    ... Experiments on the Forest Cover data set show that this parallel mixture is more accurate than a single SVM, with 90.72% accuracy reported on an independent test set. While this accuracy is impressive, the referenced paper does not consider alternative types of classifiers. In this comment, we show that a simple ensemble of decision trees results in a higher accuracy, 94.75%, and is computationally efficient. This result is somewhat surprising and illustrates the general value of experimental comparisons using different types of classifiers.

  • Learning Ensembles from Bites: a Scalable and Accurate Approach,
    Nitesh Chawla, Lawrence O. Hall, Kevin W. Bowyer and W. Philip Kegelmeyer,
    Journal of Machine Learning Research 5, April 2004, 421-451.
    pdf of this paper.
    ... Voting many classifiers built on small subsets of data is a promising approach for learning from massive data sets, one that can utilize the power of boosting and bagging. We propose a framework for building hundreds or thousands of such classifiers on small subsets of data in a distributed environment. Experiments show this approach is fast, accurate, and scalable.

  • Face Recognition Technology and the Security Versus Privacy Tradeoff,
    Kevin W. Bowyer,
    IEEE Technology and Society, Spring 2004, 9-20.
    pdf of this paper.
    Video surveillance and face recognition systems have become the subject of increased interest and controversy after the September 11 terrorist attacks on the United States. ... This paper analyzes the interplay of technical and social issues involved in the widespread application of video surveillance for person identification.
    This paper received a 2005 Award of Excellence from the Society for Technical Communication.

  • Automated Performance Evaluation of Range Image Segmentation Algorithms,
    Jaesik Min, Mark Powell, and Kevin W. Bowyer,
    IEEE Transactions on Systems, Man, and Cybernetics - Part B, 34 (1), February 2004, 263-271.
    pdf of this paper.
    ... We present an automated framework for evaluating the performance of range image segmentation algorithms. Automated tuning of algorithm parameters in this framework results in performance as good as that previously obtained with careful manual tuning by the algorithm developers. ... This framework is demonstrated using range images, but in principle it could be used to evaluate region segmentation algorithms for any type of images.

  • Assessment of Time Dependency in Face Recognition: An Initial Study,
    Patrick Flynn, Kevin W. Bowyer and P. Jonathon Phillips,
    Audio- and Video-Based Biometric Person Authentication (AVBPA 2003), Springer Lecture Notes in Computer Science 2688, 44-51.
    pdf of this paper.
    ... Experimental results suggest that (a) recognition performance is substantially poorer when unknown images are acquired on a different day from the enrolled images, (b) degradation in performance does not follow a simple predictable pattern with time between known and unknown image acquisition, and (c) performance figures quoted in the literature based on known and unknown image sets acquired on the same day may have little practical value.

  • Is Error-based Pruning Redeemable?,
    Lawrence O. Hall, Kevin W. Bowyer, Robert E. Banfield and Steven Eschrich, and Richard Collins,
    International Journal of Artificial Intelligence Tools, 12 (3), September 2003, 249-264.
    pdf of this paper.

  • Comparison and Combination of Ear and Face Images for Appearance-based Biometrics,
    Kyong Chang, Kevin W. Bowyer, Sudeep Sarkar, and Barnabas Victor,
    IEEE Transactions on Pattern Analysis and Machine Intelligence 25 (9), September 2003, 1160-1165.
    pdf of this paper.
    In the experiments reported here, recognition performance is essentially identical using ear images or face images and combining the two for multimodal recognition results in a statistically significant performance improvement. ... To our knowledge, ours is the only work to present any experimental results of computer algorithms for biometric recognition based on the ear.

  • Face Recognition Using 2D and 3D Facial Data,
    Kyong I. Chang, Kevin W. Bowyer and Patrick J. Flynn,
    First Workshop on Multi-Modal User Authentication , Santa Barbara, 25-32, December 2003.
    pdf of this paper.
    Results are presented for the largest experimental study to date that investigates the comparison and combination of 2D and 3D face recognition. To our knowledge, this is the only such study to incorporate significant time lapse between gallery and probe image acquisition ...
    Reprinted in the Journal of Intelligence Community Research and Development.

  • Distributed Learning with Bagging-like Performance,
    Nitesh Chawla, Thomas E. Moore, Lawrence O. Hall, Kevin W. Bowyer, W. Philip Kegelmeyer, and Clayton Springer,
    Pattern Recognition Letters 24 (1-3), 2003, 455-471.
    pdf of this paper.
    Bagging forms a committee of classifiers by bootstrap aggregation of training sets from a pool of training data. A simple alternative to bagging is to partition the data into disjoint subsets. Experiments with decision tree and neural network classifiers on various datasets show that, given the same size partitions and bags, disjoint partitions result in performance equivalent to, or better than, bootstrap aggregates (bags). Many applications (e.g., protein structure prediction) involve use of datasets that are too large to handle in the memory of the typical computer. Hence, bagging with samples the size of the data is impractical. Our results indicate that, in such applications, the simple approach of creating a committee of n classifiers from disjoint partitions each of size 1/n (which will be memory resident during learning) in a distributed way results in a classifier which has a bagging-like performance gain. The use of distributed disjoint partitions in learning is significantly less complex and faster than bagging.

  • SMOTE: Synthetic Minority Over-sampling TEchnique,
    Nitesh Chawla, Kevin W. Bowyer, Lawrence O. Hall, and W. Philip Kegelmeyer,
    Journal of Artificial Intelligence Research 16, 2002, 321-357.
    pdf of this paper.
    (code for example SMOTE implementation)
    This paper shows that a combination of our method of over-sampling the minority (abnormal) class and under-sampling the majority (normal) class can achieve better classifier performance (in ROC space) than only under-sampling the majority class. This paper also shows that a combination of our method of over-sampling the minority class and under-sampling the majority class can achieve better classifier performance (in ROC space) than varying the loss ratios in Ripper or class priors in Naive Bayes. Our method of over-sampling the minority class involves creating synthetic minority class examples.

  • The Gait Identification Challenge Problem: Data Sets and Baseline Algorithm,
    P. Jonathon Phillips, Sudeep Sarkar, Isidro Robledo, Patrick Grother, and Kevin W. Bowyer,
    International Conference on Pattern Recognition (ICPR 2002) , August 2002, Montreal, I:385-388.

  • ``Star Wars'' Revisited - A Continuing Case Study In Ethics and Safety-Critical Software,
    Kevin W. Bowyer,
    IEEE Technology and Society 21 (1), Spring 2002, 13-26.
    pdf of this paper.
    The Reagan-era Strategic Defense Initiative was the focus of a great deal of technical argument relating to the design and testing of safety critical software. ... This paper describes a curriculum module developed around a Reagan-era SDI debate on the theme - 'Star Wars: Can the computing requirements be met?' This module should be appropriate for use in ethics-related or software-engineering-related courses taught in undergraduate Information Systems, Information Technology, Computer Science, or Computer Engineering programs.
    This article is highlighted on the SSIT web site as one related to ABET / CSAB accreditation requirements.

  • Edge Detector Evaluation Using Empirical ROC Curves,
    Kevin W. Bowyer, Christine Kranenburg, and Sean Dougherty.
    Computer Vision and Image Understanding 84 (1), October 2001, 77-103.
    pdf of this paper.
    This paper focuses on evaluating the performance of edge detection algorithms. ... Performance is summarized using receiver operating characteristic curves. ... No other work uses adaptive parameter sampling to construct empirical ROC curves for pixel-level performance evaluation, or has used so many real images, or has compared such a broad selection of detectors.

  • Comparison of Edge Detector Performance Through Use In An Object Recognition Task,
    Min C. Shin, Dmitry B. Goldgof and Kevin W. Bowyer.
    Computer Vision and Image Understanding 84 (1), October 2001, 160-178.
    DOI link.
    Edge detector performance is measured using a particular edge-based object recognition algorithm as a higher-level task. A detector's performance is ranked according to the object recognition performance that it generates. We have used a challenging train and test dataset containing 110 images of jeep-like objects. Six edge detectors are compared and results suggest that (1) the SUSAN edge detector performs best and (2) the ranking of various edge detectors is different from that found in other evaluations.

  • Comparison of Edge Detection Algorithms Using a Structure From Motion Task,
    Min C. Shin, Dmitry B. Goldgof, Kevin W. Bowyer and S. Nikiforou,
    IEEE Transactions on Systems, Man and Cybernetics - Part B 31 (4), August 2001, 589-601.
    DOI link.
    We use the task of structure from motion as a "black box" through which to evaluate the performance of edge detection algorithms. Edge detector goodness is measured by how accurately the SFM could recover the known structure and motion from the edge detection of the image sequences. We use a variety of real image sequences with ground truth to evaluate eight different edge detectors from the literature. Our results suggest that ratings of edge detector performance based on pixel-level metrics and on the SFM are well correlated and that detectors such as the Canny detector and the Heitger detector offer the best performance.

  • Mentoring undergraduates in computer vision research,
    Mubarak Shah and Kevin W. Bowyer,
    IEEE Transactions on Education 44 (3), 252-257, August 2001.
    DOI link.
    ... During the last 14 years roughly 130 undergraduate students from several institutions have participated in the research experiences for undergraduates (REU) program funded by the National Science Foundation (NSF). A large fraction of our students have been able to prepare a paper for submission to a conference, have the paper accepted, and then attend the conference to present the paper. ...

  • Ethics and Computing: Living Responsibly In a Computerized World,
    Kevin W. Bowyer.
    IEEE Press (second edition), 2001.

  • The Digital Database for Screening Mammography,
    Michael Heath, Kevin Bowyer, Daniel Kopans, Richard Moore and W. Philip Kegelmeyer.
    in Proceedings of the Fifth International Workshop on Digital Mammography, M.J. Yaffe, ed., 212-218, Medical Physics Publishing, 2001.
    pdf of this paper.
    The Digital Database for Screening Mammography (DDSM) is a database of digitized film-screen mammograms with associated ground truth information. The purpose of this resource is to provide a large set of mammograms in a digital format that may be used by researchers to evaluate and compare the performance of computer-aided detection (CAD) algorithms. ...

  • Validation of Medical Image Analysis Techniques,
    Kevin W. Bowyer,
    chapter in Handbook of Medical Imaging: Volume 2 - Medical Image Processing and Analysis J.M. Fitzpatrick and M. Sonka, editors, SPIE Press, 2000, 567-607.
    link to copy on Google Books.

  • Resources For Teaching Ethics and Computing,
    Kevin W. Bowyer,
    Journal of Information Systems Education 11 (3-4), 91-92, Summer-Fall 2000.

  • Pornography On the Dean's PC: An Ethics and Computing Case Study,
    Kevin W. Bowyer,
    Journal of Information Systems Education 11 (3-4), 121-126, Summer-Fall 2000.
    JISE link.

  • Themes for Improved Teaching of Image Computation,
    Kevin W. Bowyer, George Stockman and Louise Stark,
    IEEE Transactions on Education 43 (2), 221-223, May 2000.
    DOI link.
    This paper reports on recommendations and results from a panel discussion and a workshop devoted to the theme of education in areas that involve image computation. One specific set of contributions is in the area of integrating image computation into required core courses such as Introduction to Computing and Data Structures. Another area of specific contributions is the development of undergraduate elective courses on topics such as robotics and medical image analysis. A third area of contributions is the improvement of traditional image processing and computer vision courses.

  • Registration and Difference Analysis of Corresponding Mammogram Images,
    Maha Y. Sallam and Kevin W. Bowyer,
    Medical Image Analysis 3 (2), 103-118, 1999.
    DOI link.
    An automated technique is proposed for identifying differences between corresponding mammogram images. The technique recovers an approximate deformation between a pair of mammograms based on identifying corresponding features across the two images. The registration process is completed using an unwarping technique for transforming one image into the coordinate system of the other. A difference image is generated using intensity-weighted subtraction in order to identify regions of large difference. Evaluation of the technique is performed using 124 bilateral image pairs which contain a total of 77 abnormalities of different types. The purpose of this paper is to measure the extent to which the mammogram registration technique is able to provide useful information for identifying abnormalities in mammograms.

  • Evaluation of Texture Segmentation Algorithms,
    Kyong Chang, Kevin W. Bowyer and S. Munishkumaran,
    Computer Vision and Pattern Recognition (CVPR '99), I:294-299.
    pdf of this paper.
    This paper presents a method of evaluating unsupervised texture segmentation algorithms. The control scheme of texture segmentation has been conceptualized as two modular processes: (1) feature computation and (2) segmentation of homogeneous regions based on the feature values. Three feature extraction methods are considered: gray level co-occurrence matrix, Laws' texture energy and Gabor multi-channel filtering. Three segmentation algorithms are considered: fuzzy c-means clustering, square-error clustering and split-and-merge. A set of 35 real scene images with manually-specified ground truth was compiled. Performance is measured against ground truth on real images using region-based and pixel-based performance metrics.

  • Dynamic-Scale Model Construction From Range Imagery,
    Adam Hoover Dmitry Goldgof, and Kevin W. Bowyer,
    IEEE Transactions on Pattern Analysis and Machine Intelligence 29 (12), 1352-1357, December 1998.
    DOI link.
    The construction of a surface model from range data may be undertaken at any point in a continuum of scales that reflects the level of detail of the resulting model. This continuum relates the construction parameters to the scale of the model. We propose methods to dynamically reprocess range data at different scales. The construction result from a single scale is automatically evaluated, causing reconstruction at a different scale when user-defined criteria are not met. We demonstrate our methods in constructing a planar b-rep space envelope (a scene representation) for over 400 range images. The experiments demonstrate that ability to construct 100 percent valid models, with the scale of detail within specified requirements.

  • Overview of Work In Empirical Evaluation of Computer Vision Algorithms,
    Kevin W. Bowyer and P. Jonathon Phillips,
    in Empirical Evaluation Techniques in Computer Vision, K.W. Bowyer and P.J. Phillips, editors, IEEE Computer Society Press, 1998, pages 1-11.
    pdf version of this chapter.

  • Function from Visual Analysis and Physical Interaction: A Methodology for Recognition of Generic Classes of Objects,
    Melanie Sutton, Louise Stark and Kevin Bowyer, Image and Vision Computing 16 (11), 745-763, August 1998.
    DOI link.
    This paper presents an overview of the GRUFF-I (Generic Recognition Using Form, Function and Interaction) system, a nonpart-based approach to generic object recognition which reasons about and generates plans for interaction with three-dimensional (3D) shapes from the categories furniture and dishes. ...

  • Current Status of the Digital Database for Screening Mammography,
    Michael Heath, Kevin Bowyer, Daniel Kopans, W. Philip Kegelmeyer, Richard Moore, Kyong Chang and S. Munishkumaran,
    in Digital Mammography, 457-460, Kluwer Academic Publishers, 1998 (proceedings of the Fourth International Workshop on Digital Mammography)
    pdf of this paper.
    The Digital Database for Screening Mammography is a resource for use by researchers investigating mammogram image analysis. In particular, the resource is focused on the context of image analysis to aid is screening for breast cancer. The database now contains substantial numbers of "normal" and "cancer" cases. ...

  • Are Edges Sufficient for Object Recognition?,
    Thomas A. Sanocki, Kevin W. Bowyer, Michael Heath, and Sudeep Sarkar,
    Journal of Experimental Psychology: Human Perception and Performance 24 (1), 340-349, January 1998.
    pdf of this paper.
    The authors argue that the concept of `edges' as used in current research on object recognition obscures the significant difficulties involved in interpreting stimulus information. ... With 1-s exposures, the accuracy of identifying objects in the edge images was found to be less than half that with color photographs. Therefore, edges are far from being sufficient for object recognition.

  • Comparison of Edge Detectors: A Methodology and Initial Study,
    Michael Heath, Sudeep Sarkar, Thomas A. Sanocki and Kevin W. Bowyer,
    Computer Vision and Image Understanding 69, (1), 38-54, January 1998.
    DOI link.
    Because of the difficulty of obtaining ground truth for real images, the traditional technique for comparing low-level vision algorithms is to present image results, side by side, and to let the reader subjectively judge the quality. This is not a scientifically satisfactory strategy. However, human rating experiments can be done in a more rigorous manner to provide useful quantitative conclusions. We present a paradigm based on experimental psychology and statistics, in which humans rate the output of low level vision algorithms. We demonstrate the proposed experimental strategy by comparing four well-known edge detectors: ...

  • A Robust Visual Method for Assessing the Relative Performance of Edge-Detection Algorithms,
    Michael Heath, Sudeep Sarkar, Thomas A. Sanocki, and Kevin W. Bowyer,
    IEEE Transactions on Pattern Analysis and Machine Intelligence 19 (12), 1338-1359, December 1997.
    pdf of this paper.
    A new method for evaluating edge detection is presented ... The basic measure of performance is a visual rating score which indicates the perceived quality of the edges for identifying an object. ... The novel aspect of this work is the use of a visual task and real images of complex scenes in evaluating edge detectors.

  • Generating ROC curves for artificial neural networks,
    Kevin S. Woods and Kevin W. Bowyer,
    IEEE Transactions on Medical Imaging 16 (3), 329-337, June 1997.
    DOI link.
    Receiver operating characteristic (ROC) analysis is an established method of measuring diagnostic performance in medical imaging studies. Traditionally, artificial neural networks (ANN's) have been applied as a classifier to find one "best" detection rate. Recently researchers have begun to report ROC curve results for ANN classifiers. The current standard method of generating ROC curves for an ANN is to vary the output node threshold for classification. Here, the authors propose a different technique for generating ROC curves for a two class ANN classifier. They show that this new technique generates better ROC curves in the sense of having greater area under the ROC curve (AUC), and in the sense of being composed of a better distribution of operating points.

  • Combination of Multiple Classifiers Using Local Accuracy Estimates,
    Kevin S. Woods, W. Philip Kegelmeyer and Kevin W. Bowyer,
    IEEE Transactions on Pattern Analysis and Machine Intelligence 19 (4), 405-410, April 1997.
    pdf of this paper.
    This paper presents a method for combining classifiers that uses estimates of each individual classifier's local accuracy in small regions of the feature space surrounding an unknown sample. An empirical evaluation using five real data sets confirms the validity of our approach compared to some other combination of multiple classifiers algorithms. We also suggest a methodology for determining the best mix of individual classifiers.

  • Recognizing object function through reasoning about partial shape descriptions and dynamic physical properties,
    Louise Stark, Kevin W. Bowyer, Adam W. Hoover and Dmitry B. Goldgof,
    Proceedings of the IEEE 84 (11), 1640-1656, November 1996.
    DOI link.
    Knowledge about required functionality of an object can be used as an effective representation for a generic object category (e.g., "chair", "cup", or "hammer"). This approach to object representation and recognition has recently become an active area of research. We explore a scenario in which a robot senses the environment to obtain an initial partial shape model of an object. If the information in this initial model is not sufficient to hypothesize a possible function for the object, then additional view(s) may be suggested. Once a possible function is hypothesized, a plan is formulated for interacting with the object to confirm that its material properties are compatible with the hypothesized function. The module for reasoning about partial shape models has been evaluated on over 200 shape models acquired from range images. The module for carrying out a function verification plan has been evaluated in a simulated environment using the ThingWorld (TW) system.

  • An Experimental Comparison of Range Image Segmentation Algorithms,
    Adam W. Hoover, Gillian Jean-Baptiste, Xiaoyi Jiang, Patrick Flynn, Horst Bunke, Dmitry Goldgof, Kevin W. Bowyer, David Eggert, Andrew Fitzgibbon, and Robert Fisher,
    IEEE Transactions on Pattern Analysis and Machine Intelligence 18, (7), 673-689, July 1996.
    pdf of this paper.
    This paper evaluates four segmentation algorithms on 80 real images with ground truth and objective performance measures. ... This type of framework for a competitive effort is essentially never used in mainstream computer vision, though it is standard practice in some related areas ... Beside the development of a philosophy of comparative experiment research, an important contribution here is an assessment of the state-of-the-art in planar range image segmentation. Based on our results, we assert that this problem is not 'solved.' This finding may be surprising and possibly controversial. We would welcome an empirical demonstration that the claim is false.

  • Learning Membership Functions in a Function-based Object Recognition System,
    Kevin S. Woods, Diane Cook, Lawrence Hall, Louise Stark and Kevin W. Bowyer,
    Journal of Artificial Intelligence Research 3, 187-222, October 1995.

  • On recovering hyperquadrics from range data,
    Senthil Kumar, Song Han, Dmitry Goldgof, and Kevin W. Bowyer,
    IEEE Transactions on Pattern Analysis and Machine Intelligence 17, (11), 1079-1083, November 1995. Computer Vision and Image Understanding 62 (2), 177-193, September 1995.

  • Extracting a Valid Boundary Representation From a Segmented Range Image,
    Adam W. Hoover, Dmitry Goldgof, and Kevin W. Bowyer,
    IEEE Transactions on Pattern Analysis and Machine Intelligence 17, (9), 920-924, September 1995.
    A new approach is presented for extracting an explicit 3-D shape model from a single range image. One novel aspect is that the model represents both observed object surfaces, and surfaces which bound the volume of occluded space. Another novel aspect is that the approach does not require that the range image segmentation be perfect. ... A third novel aspect of this work is that the implementation has been evaluated on over 200 real images of polyhedral objects with no operator intervention and all parameters held constant, and obtained a 97% success rate in creating valid b-reps.
    DOI link.

  • Generic Recognition of Articulated Objects Through Reasoning About Potential Function,
    Kevin Green, David Eggert, Louise Stark, and Kevin W. Bowyer,
    Computer Vision and Image Understanding 62 (2), 177-193, September 1995.

  • Aspect Graphs and Their Use in Object Recognition,
    David Eggert, Louise Stark, and Kevin W. Bowyer,
    Annals of Mathematics and Artificial Intelligence 13, 347-375, 1995.

  • GRUFF-3: Generalizing the Domain of a Function-based Recognition System,
    Melanie Sutton, Louise Stark, and Kevin W. Bowyer, Pattern Recognition 27 (12), 1743-1766, December 1994.

  • Function-based Generic Recognition for Multiple Object Categories,
    Louise Stark and Kevin W. Bowyer,
    CVGIP: Image Understanding 59 (1), 1-21, January 1994.
    ... Our work concentrates specifically on the relation between shape and function of rigid 3D objects. Recognition of an observed shape is performed by reasoning about the function that it might serve. Previous efforts have dealt with only a single basic level object category. A number of important issues arise in extending this approach to deal with multiple basic-level categories. ...

  • Comparative evaluation of pattern recognition techniques for detection of microcalcifications in mammography,
    Kevin S. Woods, Christopher C. Doss, Kevin W. Bowyer, Jeffrey L. Solka, Carey E. Priebe, and W. Philip Kegelmeyer,
    International Journal of Pattern Recognition and Artificial Intelligence 7 (6), 1417-1436, December 1993.
    Computer-assisted detection of microcalcifications in mammographic images will likely require a multistage algorithm that includes segmentation of possible microcalcifications, pattern recognition techniques to classify the segmented objects, a method to determine if a cluster of calcifications exists, and possibly a method to determine the probability of malignancy. This paper focuses on the first three of these stages ...

  • The Scale Space Aspect Graph,
    David W. Eggert, Kevin W. Bowyer, Charles R. Dyer, Henrik I. Christensen, and Dmitry B. Goldgof,
    IEEE Transactions on Pattern Analysis and Machine Intelligence 15 (11), 1114-1130, November 1993.
    pdf of this paper.
    This paper introduces the concept of the scale space aspect graph, defines three different interpretations of the scale dimension, and presents a detailed example for a simple class of objects, with scale defined in terms of the spatial extent of features in the image.

  • An Investigation of Methods of Combining Functional Evidence for 3-D Object Recognition,
    Louise Stark, Lawrence O. Hall, and Kevin W. Bowyer,
    International Journal of Pattern Recognition and Artificial Intelligence 7 (3), 573-594, June 1993.

  • Computing the Generalized Aspect Graph for Objects with Moving Parts,
    Kevin W. Bowyer, Maha Y. Sallam, David Eggert, and John H. Stewman,
    IEEE Transactions on Pattern Analysis and Machine Intelligence 15 (6), 605-610, June 1993.
    A number of researchers have described algorithms for computing the aspect graph representation, but this work has thus far been limited to entirely rigid objects. We generalize this concept to include a larger, more realistic domain of objects known as articulated assemblies, those objects composed of rigid parts with articulated connections allowed between parts.

  • Computing the Perspective Projection Aspect Graph of Solids of Revolution,
    David W. Eggert and Kevin W. Bowyer,
    IEEE Transactions on Pattern Analysis and Machine Intelligence 15 (2), 109-128, February 1993.
    This paper presents the first (only) implemented algorithm to compute the aspect graph for a class of curved-surface objects based on an exact parcellation of 3-D viewpoint space. The class of objects considered is solids of revolution. ... the worst-case complexity of ... the number of cells in the aspect graph is is O(N^4), where N is the degree of a polynomial that defines the object shape.

  • Achieving Generalized Object Recognition Through Reasoning About Association of Function To Structure,
    Louise Stark and Kevin W. Bowyer,
    IEEE Transactions on Pattern Analysis and Machine Intelligence 13 (10), 1097-1104, October 1991.
    pdf of this paper.
    The purpose of the work described here is to demonstrate the feasibility of defining an object category in terms of the functional properties shared by all objects in the category. ... A complete system has been implemented that takes the boundary surface description of a 3-D object as its input and attempts to recognize whether the object belongs to the category chair ... This is, to our knowledge the first (only) implemented system to explore the use of a purely function-based definition of an object category ... to recognize 3-D objects.

  • Aspect Graphs: An Introduction and Survey of Recent Results,
    Kevin W. Bowyer and Charles R. Dyer,
    International Journal of Imaging Systems and Technologies 2, 315-328, 1990.

  • Direct Construction of Perspective Projection Aspect Graphs for Planar-Face Convex Objects,
    John Stewman and Kevin W. Bowyer,
    Computer Vision, Graphics and Image Processing 51 (1), 20-37, July 1990.

  • The Effects of Age on Radionuclide Angiographic Detection and Quantitation of Left-to-right Shunts,
    Page A.W. Anderson, Kevin W. Bowyer, and Robert H. Jones,
    American Journal of Cardiology 53, 879-883, March 1984.

  • Optimizing Contiguous Element Region Selection for Virtual Memory Computer Systems,
    Kevin W. Bowyer, and C. Frank Starmer,
    IEEE Transactions on Computers 32, 1201-1203, December 1983.

  • Radionuclide Analysis of Right and Left Ventricular Response to Exercise in Patients with Atrial and Ventricular Septal Defects,
    Claude A. Peter, Kevin W. Bowyer, and Robert H. Jones,
    American Heart Journal 105, 428-435, March 1983.

  • Error Sensitivity of Computed Tomography Guided Stereotaxis,
    Kevin W. Bowyer, C. Frank Starmer and P. J. DuBois,
    Computers and Biomedical Research 15, 272-280, 1982.

  • A Simulation-based Sensitivity Study of Angiocardiographic Approaches to Shunt Assessment,
    Kevin W. Bowyer and C. Frank Starmer,
    Computers and Biomedical Research 15, 111-128, 1982.

  • CT-guided Stereotaxis Using a Modified Conventional Stereotaxic Frame,
    P. J. DuBois, B. S. Nashold, J. Perry, P. Burger, Kevin W. Bowyer, E. R. Heinz, B. P. Drayer, S. Bigner, and A. C. Higgins,
    American Journal of Neuroradiology 3, 345-351, 1982.

Home

Research

Teaching

Professional

Personal