Peter Wild

Weighted adaptive Hough and ellipsopolar transforms for real-time iris segmentation

Efficient and robust segmentation of less intrusively or non-cooperatively captured iris images is still a challenging task in iris biometrics. This paper proposes a novel two-stage algorithm for the localization and mapping of iris texture in images of the human eye into Daugmans doubly dimensionless polar coordinates. Motivated by the growing demand for real-time capable solutions, coarse center detection and fine boundary localization usually combined in traditional approaches are decoupled. Therefore, search space at each stage is reduced without having to stick to simpler models. Another motivation of this work is independence of sensors. A comparison of reference software on different datasets highlights the problem of database-specific optimizations in existing solutions. This paper instead proposes the application of Gaussian weighting functions to incorporate model-specific prior knowledge. An adaptive Hough transform is applied at multiple resolutions to estimate the approximate position of the iris center. Subsequent polar transform detects the first elliptic limbic or pupillary boundary, and an ellipsopolar transform finds the second boundary based on the outcome of the first. This way, both iris images with clear limbic (typical for visible-wavelength) and with clear pupillary boundaries (typical for near infrared) can be processed in a uniform manner.

A Ground Truth for Iris Segmentation

Classical iris biometric systems assume ideal environmental conditions and cooperative users for image acquisition. When conditions are less ideal or users are uncooperative or unaware of their biometrics being taken the image acquisition quality suffers. This makes it harder for iris localization and segmentation algorithms to properly segment the acquired image into iris and non-iris parts. Segmentation is a critical part in iris recognition systems, since errors in this initial stage are propagated to subsequent processing stages. Therefore, the performance of iris segmentation algorithms is paramount to the performance of the overall system. In order to properly evaluate and develop iris segmentation algorithm, especially under difficult conditions like off angle and significant occlusions or bad lighting, it is beneficial to directly assess the segmentation algorithm. Currently, when evaluating the performance of iris segmentation algorithms this is mostly done by utilizing the recognition rate, and consequently the overall performance of the biometric system. In order to streamline the development and assessment of iris segmentation algorithms with the dependence on the whole biometric system we have generated a iris segmentation ground truth database. We will show a method for evaluating iris segmentation performance base on this ground truth database and give examples of how to identify problematic cases in order to further analyse the segmentation algorithms.

Reliability-balanced feature level fusion for fuzzy commitment scheme

Fuzzy commitment schemes have been established as a reliable means of binding cryptographic keys to binary feature vectors extracted from diverse biometric modalities. In addition, attempts have been made to extend fuzzy commitment schemes to incorporate multiple biometric feature vectors. Within these schemes potential improvements through feature level fusion are commonly neglected. In this paper a feature level fusion technique for fuzzy commitment schemes is presented. The proposed reliability-balanced feature level fusion is designed to re-arrange and combine two binary biometric templates in a way that error correction capacities are exploited more effectively within a fuzzy commitment scheme yielding improvement with respect to key-retrieval rates. In experiments, which are carried out on iris-biometric data, reliability-balanced feature level fusion significantly outperforms conventional approaches to multi-biometric fuzzy commitment schemes confirming the soundness of the proposed technique.

Incremental iris recognition: A single-algorithm serial fusion strategy to optimize time complexity

Daugmans algorithm, mapping iris images to binary codes and estimating similarity between codes applying the fractional Hamming Distance, forms the basis of todays commercially used iris recognition systems. However, when applied to large-scale databases, the linear matching of a single extracted iris-code against a gallery of templates is very time consuming and a bottleneck of current implementations. As an alternative to pre-screening techniques, our work is the first to present an incremental approach to iris recognition. We combine concentration of information in the first bits of an iris-code with early rejection of unlikely matches during matching stage to incrementally determine the best-matching candidate in the gallery. Our approach can transparently be applied to any iris-code based system and is able to reduce bit comparisons significantly (to about 5% of iris-code bits) while exhibiting a Rank-1 Recognition Rate being at least as high as for matches involving all bits.

Multi-stage visible wavelength and near infrared iris segmentation framework

This paper presents a multi-stage iris segmentation framework for the localization of pupillary and limbic boundaries of human eyes. Instead of applying time-consuming exhaustive search approaches, like traditional circular Hough Transform or Daugmans integrodifferential operator, an iterative approach is used. By decoupling coarse center detection and fine boundary localization, faster processing and modular design can be achieved. This alleviates more sophisticated quality control and feedback during the segmentation process. By avoiding database-specific optimizations, this work aims at supporting different sensors and light spectra, i.e. Visible Wavelength and Near Infrared, without parameter tuning. The system is evaluated by using multiple open iris databases and it is compared to existing classical approaches.

Single-sensor multi-instance fingerprint and eigenfinger recognition using (weighted) score combination methods

When multiple instances of single biometrics can be acquired from a single input simultaneously, a multiple-step acquisition at additional transaction time cost can be avoided. We present a rotation-invariant, peg-free multi-instance fingerprint- and eigenfinger-based biometric system extracting multiple features from a palmar scan of the hand. Our evaluation targets: (1) rankings of individual fingers with respect to minutiae and eigenfinger features; (2) fusion of multi-instance intra-feature (minutiae or eigenfinger) matching scores; (3) cross-feature compared to intra-feature performance; (4) optimal weights for weighted versions of five score-level fusion methods – max, median, min, product and sum and (5) aspects of computational demands for hand-based identification discussing the usage of serial classifier combinations instead of classically employed parallel ones. We examine results of an experimental approach to the problem of finding a suitable fusion method by investigating the effect of matcher-specific combination weights on recognition accuracy and compare cross-feature and intra-feature score combinations.

Shifting score fusion: on exploiting shifting variation in iris recognition

Iris recognition applies pattern matching techniques to compare two iris images and retrieve a comparison score that reflects their degree of (dis-)similarity. While numerous approaches to generating iris-codes have been proposed for the relatively young discipline of automated iris recognition, there are only few, usually simple, comparison techniques, e.g. fractional Hamming distance. However, in case of having access to specific iris-codes only or black-boxed feature extraction, there may be situations where improved comparison (even at potentially higher processing cost) is desirable. In this paper we present a new strategy for comparing iris-codes, which utilizes variations within comparison scores at different shift positions. We demonstrate that by taking advantage of this information, which even comes at negligible cost, recognition performance is significantly improved. The soundness of the approach is confirmed by experiments using two different iris-code based feature extraction algorithms.

On combining selective best bits of iris-codes

This paper describes a generic fusion technique for iris recognition at bit-level we refer to as Selective Bits Fusion. Instead of storing multiple biometric templates for each algorithm, the proposed approach extracts most discriminative bits from multiple algorithms into a new template being even smaller than templates for individual algorithms. Experiments for three individual iris recognition algorithms on the open CASIA-V3-Interval iris database illustrate the ability of this technique to improve accuracy and processing time simultaneously. In all tested configurations Selective Bits Fusion turned out to be more accurate than fusion using the Sum Rule while being about twice as fast. The design of the new template allows explicit control of processing time requirements and introduces a tradeoff between time and accuracy of biometric fusion, which is highlighted in this work.

Footprint-based biometric verification

We investigate the potential of foot biometric features based on geometry, shape, and texture and present algorithms for a prototype rotation invariant verification system. An introduction to origins and fields of application for footprint-based personal recognition is accompanied by a comparison with traditional hand biometry systems. Image enhancement and feature extraction steps emphasizing specific characteristics of foot geometry and their permanence and distinctiveness properties, respectively, are discussed. Collectability and universality issues are considered as well. A visualization of various test results comparing discriminative power of foot shape and texture is given. The impact on real-world scenarios is pointed out, and a summary of results is presented.

Enhancing iris matching using levenshtein distance with alignment constraints

Iris recognition from surveillance-type imagery is an active research topic in biometrics. However, iris identification in unconstrained conditions raises many proplems related to localization and alignment, and typically leads to degraded recognition rates. While development has mainly focused on more robust preprocessing, this work highlights the possibility to account for distortions at matching stage. We propose a constrained version of the Levenshtein Distance (LD) for matching of binary iris-codes as an alternative to the widely accepted Hamming Distance (HD) to account for iris texture distortions by e.g. segmentation errors or pupil dilation. Constrained LD will be shown to outperform HD-based matching on CASIA (third version) and ICE (2005 edition) datasets. By introducing LD alignment constraints, the matching problem can be solved in O(nċ s) time and O(n+s) space with n and s being the number of bits and shifts, respectively.