Overview of the multiple biometrics grand challenge

Abstract

Recent studies show that face recognition in uncontrolled images remains a challenging problem, although the reasons why are less clear. Changes in illumination are one possible explanation, although algorithms developed since the advent of the PIE and Yale B data bases supposedly compensate for illumination variation. Edge density has also been shown to be a strong predictor of algorithm failure on the FRVT 2006 uncontrolled images: recognition is harder on images with higher edge density. This paper presents a new study that explains the edge density effect in terms of illumination and shows that top performing algorithms in FRVT 2006 are still sensitive to lighting. This new study also shows that focus, originally suggested as an explanation for the edge density effect, is not a significant factor. The new lighting model developed in this study can be used as a measure of face image quality.

Iris recognition by fusing different representations of multi-scale Taylor expansion

“…We use three publicly available iris databases in all experiments: the Chinese Academy of Sciences Casia 2.0 (device 1) [31], the US National Institute of Standards and Technology (NIST) ''Iris Challenge Evaluation'', experiment 1 (right eye), (ICE-1) [32], and the NIST ''Multiple Biometric Grand Challenge'' (MBGC), Portal Challenge, experiment 3, left eye (MBGC-3l) [33]. Their properties are summarized in Table 3.…”

Dictionary-Based Face Recognition from Video

“…To illustrate the effectiveness of our method, we present experimental results on three publicly available datasets for video-based face recognition: the Multiple Biomertic Grand Challenge (MBGC) [18], [19], the Face and Ocular Challenge Series (FOCS) [20], [21], and the Honda/UCSD datasets [7]. Figure 2(a) shows example frames from four different activity sequences, where each subject reads from a paper, and the sequences consists of non-frontal views of the subject.…”

“…In the MBGC [18] protocol, verifications are specified by two sets: target and query. The protocol requires the algorithm to match each target sequence with all query sequences.…”

Appearance-based periocular features in the context of face and non-ideal iris recognition

“…Similarly for the right eye, the Rank-1 [26] recognition rates for the iris, periocular regions, and their combination were 10.1, 88.7, and 92.4%, respectively. Even though higher iris recognition performance was reported for the MBGC portal chalenge [19], the target images used in these experiments were still images of relatively higher quality (and of a significatly larger size). In this work, both the target and the query images are from the NIR face videos and as described earlier, of highly non-ideal in nature.…”

“…Alternatively, robust algorithms can be developed to handle specific challenges such as pose or illumination invariant face recognition, or iris recognition in off-axis images. Another option is to make use of additional information to enhance the recognition performance such as the use of multi-biometrics, i.e., fusion of different biometric modalities (for example, fusing face and iris recognition scores [19]). A related idea is to explore newer traits that can aid the existing biometric modalities.…”