Iris Recognition on Mobile Devices Using Near-Infrared Images

Haiqing Li , ... Zhenan Sun , in Human Recognition in Unconstrained Environments, 2017

5.2 Preprocessing

Iris images acquired by mobile devices usually contain not only periocular regions but also partial face regions. As shown in Fig. 5.4, the major task for image preprocessing is to detect eye regions and then isolate the valid iris regions from the background. Cho et al. [3,4] are among the first researchers to investigate the iris segmentation algorithms specifically for mobile phones. The intensity characteristics of iris images are exploited to design real-time rule-based algorithms. In addition, floating point operations which are time-consuming on ARM CPU are removed to reduce the processing time.

Figure 5.4

Figure 5.4. Image preprocessing.

Although rule-based iris detection and segmentation methods are fast, they cannot deal with low quality iris images. Since the computational capability of mobile devices has been improved greatly, more complex preprocessing algorithms can be utilized. For example, periocular regions are first localized by Adaboost eye detectors [5]. Then, the inner and outer iris boundaries and eyelids are localized by integro-differential operators [6] or Hough transforms [7]. Thirdly, horizontal rank filtering and histogram filtering can be successively used for eyelash and shadow removal [8]. Finally, the isolated iris texture is unfolded to a rectangle image by the homogeneous rubber sheet model [6].

To solve the problem of low resolution iris images acquired by mobile devices, a straightforward idea is to increase the resolution of iris images. Super-resolution (SR) is widely used to increase image resolution. It usually takes one or more low resolution (LR) images as input and maps them to a high resolution (HR) output image. Single image super-resolution (SISR) is a popular research topic nowadays. SR in many computer vision tasks only focuses on visual effect [9], while SR in biometrics mainly aims at improving the recognition rate [10]. After SR, higher resolution iris images or enhanced feature codes are fed into the traditional recognition procedure. In this way, the recognition accuracy is expected to be improved.

We evaluated two pixel level SISR methods which were proposed recently. The first one is Super-Resolution Convolutional Neural Networks (SRCNN) [11]. It learns the nonlinear mapping function between LR images and HR images. The convolutional neural networks (CNNs) have a lightweight structure that only has three convolutional layers, as shown in Fig. 5.5. The loss function is computed as the mean squared error between the reconstructed images and the corresponding ground-truth HR images. It takes three days to train a SRCNN model using 91 images on a GTX 770 GPU. The second method is Super-Resolution Forests (SRF) [12]. Random forests have merits of being highly nonlinear, and are usually extremely fast during both the training and evaluation phases. SRF build on linear prediction models in leaf nodes. During tree growing, a novel regularized objective function is adopted that operates on both output and input domains. SRF can be trained within minutes on a single CPU core, which is very efficient. The SR models for iris recognition are trained by HR images acquired by IrisGuard and the corresponding downsampled LR images. At the testing stage, we input one normalized LR iris image into the trained model and the corresponding HR iris image is output.

Figure 5.5

Figure 5.5. The SRCNN model with one normalized iris image as input.

Two mobile iris databases are used to evaluate the effectiveness of the above two SISR methods in improving the recognition rate. The first database is the CASIA-Iris-Mobile-V1.0 that includes 2800 iris images from 70 Asians. The second database is CASIA-Iris-Mobile-V2.0 that contains 12,000 iris images from 200 Asians. After super-resolution of LR normalized images, we extract Ordinal Measures (OMs) [16] features from HR images for recognition. Receiver operating characteristic (ROC) curves on the first and second databases are shown in Fig. 5.6 and Fig. 5.7, respectively.

Figure 5.6

Figure 5.6. ROC curves of SRCNN and SRF on CASIA-Iris-Mobile-V1.0 database.

Figure 5.7

Figure 5.7. ROC curves of SRCNN and SRF on CASIA-Iris-Mobile-V2.0 database.

Experiments on these two databases get similar conclusions: (i) the SRCNN and SRF methods get comparable recognition results. SRCNN takes about 3 s on a single normalized image with size of 70 × 540 while the SRF takes only about 0.3 s on the same image. The SRF is much faster; (ii) SISR has limited effectiveness in improving the recognition accuracy. The limitations are as follows: pixel level SISR is not directly related to recognition and may introduce artifacts; the SR model is trained with synthesized LR images that are very different from real-world LR images. We need to focus attention on how to access more information, e.g., by adopting multi-frame SR that can use complementary information from different frames.

In order to directly boost the recognition accuracy, SR can be applied at the feature and code level. Nguyen et al. [13] propose a novel feature-domain SR approach using 2D Gabor wavelets. The SR output (a super-resolved feature vector) is directly employed for recognition. Liu et al. [14] propose a code-level scheme for heterogeneous matching of LR and HR iris images. They use an adapted Markov network to establish the statistical relationship between a number of binary codes of LR iris images and a binary code corresponding to the latent HR iris image. Besides, the co-occurrence relationship between neighboring bits of HR iris code is also modeled through this Markov network. Therefore, an enhanced iris feature code from the probe set of LR iris image sequences can be obtained. Both of the above SR methods can achieve improved performance compared to pixel level SR.

Read full chapter

URL:

https://www.sciencedirect.com/science/article/pii/B9780081007051000051

How Iris Recognition Works

John Daugman , in Handbook of Image and Video Processing (Second Edition), 2005

9 Appendix: Two-Dimensional Focus Assessment at the Video Frame Rate

The acquisition of iris images in good focus is made difficult by the optical magnification requirements, the restrictions on illumination, and the target motion, distance, and size. All of these factors act to limit the possible depth of field of the optics, because they create a requirement for a lower F number to accommodate both the shorter integration time (to reduce motion blur) and the light dilution associated with long focal length. The iris is a 1-cm target within a roughly 3-cm-wide field that one would like to acquire at a range of about 30 cm to 50 cm, and with a resolution of about five line pairs per mm. In a fixed-focus optical system, the acquisition of iris images almost always begins in poor focus. It is therefore desirable to compute focus scores for image frames very rapidly, either to control a moving lens element or to provide audible feedback to the subject for range adjustment, or to select which of several frames in a video sequence is in best focus.

Optical defocus can be fully described as a phenomenon of the 2D Fourier domain. An image represented as a 2D function of the real plane, I(x, y), has a 2D Fourier transform F(μ, v) defined as:

(15) F ( μ , v ) = 1 ( 2 π ) 2 ∫ ∫ I ( x , y ) exp ( − i ( μ x + v y ) ) d x d y

In the image domain, defocus is normally represented as convolution of a perfectly focused image by the 2D point-spread function of the defocused optics. This point-spread function is often modelled as a Gaussian whose space constant is proportional to the degree of defocus. Thus, for perfectly focused optics, this optical point-spread function shrinks almost to a delta function, and convolution with a delta function has no effect on the image. Progressively defocused optics equates to convolving with ever wider point-spread functions.

If the convolving optical point-spread function causing defocus is an isotropic Gaussian whose width represents the degree of defocus, it is clear that defocus is equivalent to multiplying the 2D Fourier transform of a perfectly focused image with the 2D Fourier transform of the "defocusing" (convolving) Gaussian. This latter quantity is itself just another 2D Gaussian within the Fourier domain, and its spread constant there (σ) is the reciprocal of that of the image-domain convolving Gaussian that represented the optical point-spread function. Thus the 2D Fourier transform Dσ (μ, v) of an image defocused by degree 1/σ can be related to F(β, v), the 2D Fourier transform of the corresponding perfectly focused image, by a simple model such as:

(16) D σ ( μ , v ) = exp ( μ 2 − v 2 σ ) F ( μ , v )

his expression reveals that the effect of defocus is to attenuate primarily the highest frequencies in the image and that lower frequency components are affected correspondingly less, since the exponential term approaches unity as the frequencies (μ, v) become small. (For simplicity, this analysis has assumed isotropic optics and isotropic blur, and the optical point-spread function has been described as a Gaussian just for illustration. But the analysis can readily be generalized to non-Gaussian and to anisotropic optical point-spread functions.)

This spectral analysis of defocus suggests that an effective way to estimate the quality of focus of a broadband image is simply to measure its total power in the 2D Fourier domain at higher spatial frequencies, since these are the most attenuated by defocus. One may also perform a kind of "contrast normalization" to make such a spectrally-based focus measure independent of image content, by comparing the ratio of power in higher frequency bands to that in slightly lower frequency bands. Such spectrally based measurements are facilitated by exploiting Parseval's theorem for conserved total power in the two domains:

(17) ∫ ∫ | I ( x , y ) | 2 d x d y = ∫ ∫ | F ( μ , v ) | 2 d μ d v

Thus, high-pass filtering an image, or bandpass filtering it within a ring of high spatial frequency (requiring only a 2D convolution in the image domain), and integrating the power contained in it, is equivalent to computing the actual 2D Fourier transform of the image (a more costly operation) and performing the corresponding explicit measurement in the selected frequency band. Since the computational complexity of a fast Fourier transform on n × n data is O(n 2 log2 n), some 3 million floating-point operations are avoided which would be otherwise be needed to compute the spectral measurements explicitly. Instead, only about 6,000 integer multiplications per image are needed by this algorithm, and no floating-point operations. Computation of focus scores is based only on simple algebraic combinations of pixel values within local closed neighborhoods, repeated across the image.

Pixels are combined according to the following (8 × 8) convolution kernel:

The simple weights mean that the sum of the central (4 × 4) pixels can just be tripled, and then the outer 48 pixels subtracted from this quantity; the result is squared and accumulated as per (17); and then the kernel moves to the next position in the image, selecting every fourth row and fourth column. This highly efficient discrete convolution has a simple 2D Fourier analysis.

The kernel shown is equivalent to the superposition of two centered square box functions, one of size (8 × 8) and amplitude −1, and the other one of size (4 × 4) and amplitude +4. (For the central region in which they overlap, the two therefore sum to +3.) The 2D Fourier transform of each of these square functions is a 2D "sinc" function, whose size parameters differ by a factor of two in each of the dimensions and whose amplitudes are equal but opposite, since the two component boxes have equal but opposite volumes. Thus the overall kernel has a 2D Fourier transform K(μ, v)which is the difference of two, differently-sized, 2D sinc functions:

(18) K ( μ , v ) = sin ( μ ) sin ( v ) π 2 μ v − sin ( 2 μ ) sin ( 2 v ) 4 π 2 μ v

The square of this function of μ and v in the 2D Fourier domain is plotted in Fig. 11, revealing K 2(μ, v), the convolution kernel's 2D power spectrum.

FIGURE 11. The two-dimensional Fourier power spectrum of the convolution kernel used for rapid focus assessment.

Clearly, low spatial frequencies (near the center of the power spectral plot in Fig. 11) are ignored, reflecting the fact that the pixel weights in the convolution kernel all sum to zero, while a bandpass ring of upper frequencies are selected by this filter. The total power in that band is the spectral measurement of focus. Finally, this summated 2D spectral power is passed through a compressive nonlinearity of the form: f(x) = 100 · x 2/(x 2 + c 2) (where parameter c is the half-power corresponding to a focus score of 50%), to generate a normalized focus score in the range of 0 to 100 for any image. The complete execution time of this 2D focus assessment algorithm, implemented in C using pointer arithmetic, operating on a (480 × 640) image, is 15 msec on a 300-MHz RISC processor.

Read full chapter

URL:

https://www.sciencedirect.com/science/article/pii/B9780121197926501339

A Multiscale Sequential Fusion Approach for Handling Pupil Dilation in Iris Recognition

Raghunandan Pasula , ... Arun Ross , in Human Recognition in Unconstrained Environments, 2017

4.1.1 Pupil Dilation

There are several factors that negatively impact the matching accuracy of an iris recognition system. These include out-of-focus imaging, occlusions, specular reflections, image blur, non-uniform illumination, sensor noise, and pupil dilation. While most of these factors are a result of human–sensor interaction, pupil dilation is a result of ambient factors such as intensity of visible light entering the eye, psychological factors such as stress, and chemical factors such as alcohol intake [1] and drugs. Eye drops can also regulate the size of the pupil. Pupil dilation is measured in terms of pupil dilation ratio (ρ) which is the ratio of pupil radius over iris radius:

ρ = pupil radius iris radius .

ρ typically varies from 0.2 to 0.8 in humans [5]. Smaller values of ρ indicate pupil constriction while larger values of ρ indicate pupil dilation. Fig. 4.3 shows two iris images of the same person at different visible illumination levels with a large difference in pupil dilation ratio.

Figure 4.3

Figure 4.3. Images of the same eye with different pupil sizes. Iris texture undergoes complex deformation when the pupil size changes. When the pupil dilates, the iris constricts, and vice-versa.

Traditionally, the region between the pupillary and limbic boundaries is linearly sampled in order to unwrap it into a rectangular grid. This is referred to as the rubber sheet model [5]. It is often assumed, for the purpose of linear sampling, that the pupillary and limbic boundary are circular. One implementation of the rubber sheet model samples pixels radially along a fixed angular direction at regular intervals and maps them into a single column in the normalized image. This is repeated across multiple angular directions corresponding to different angular values. It was previously assumed that this type of linear sampling would be sufficient to handle changes in iris due to variation in pupil size. However, as can be seen in Fig. 4.4, employing the linear sampling of images in Fig. 4.3 results in different outputs.

Figure 4.4

Figure 4.4. Normalized images based on linear sampling of the images in Fig. 4.3A and B. It is observed that the normalized images do not register well in case of large differences in pupil size. Collarette is manually annotated in yellow to highlight the non-correspondence of iris pixels in the normalized images. Hamming distance between these two images is 0.3244 which is very close to being considered a non-match. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this chapter.)

The iris texture undergoes complex 3-D deformation when the pupil dilates or constricts. Hence, the linear normalization technique employed in Daugman [5] may not be a suitable method to match two images of the same iris with substantially different pupil sizes [7].

There are multiple ways to address this problem. These are pictorially represented in Fig. 4.5 and briefly discussed below.

Figure 4.5

Figure 4.5. Flow chart showing different stages in the iris matching pipeline at which the pupil dilation problem may be addressed. Normalized images have been enhanced for better visualization.

Solutions to the problem of matching iris images with disparate pupil size may be incorporated at different stages of an iris recognition system, viz., image acquisition, normalization, encoding, and matching. Each of these stages could be modified either independently or in conjunction with the other stages to improve the iris matching performance.

Image Prediction.

A deformation model based on the biology of the iris and the pupil could be used to predict how the iris texture would change as a function of pupil radius. Once the deformed iris pattern is predicted, subsequent processing steps such as normalization, encoding and matching can be applied without any modification. Examples of such an approach can be found in Wyatt [17] and Clark et al. [2].

Non-linear Sampling.

A typical normalization technique radially samples the iris linearly in each angular position. The sampling scheme can be modified to better match the two normalized images [15]. Image prediction models could be potentially used to guide the non-linear sampling method.

New Encoding Method.

A new feature encoding method can be developed to better match iris images with disparate pupil sizes. A corresponding matching method may have to be developed depending on the type of extracted features.

In this work, we propose a multi-scale fusion scheme where the IrisCodes generated using multiple Gabor filters are sequentially fused at the bit-level. The proposed scheme and its rationale is described in detail in Section 4.5.

Read full chapter

URL:

https://www.sciencedirect.com/science/article/pii/B978008100705100004X

Optimization of Methods for Image-Texture Segmentation Using Ant Colony Optimization

Manasa Nadipally , in Intelligent Data Analysis for Biomedical Applications, 2019

2.5.4 Ant Colony Optimization-Image Segmentation on Iris Images

The performance of an ACO-based image segmentation algorithm for iris images is demonstrated in Fig. 2.6. It can be easily observed that the pigmentation and collarette regions are distinct under the ACO-ISA based detector. It also establishes that the ACO-ISA processed image depicts better quality for further application of region edging techniques.

Figure 2.6. (A) Original iris image. (B) Performance of Sobel detector (at 200 iterations). (C) Performance of Canny detector at 200 iterations. (D) Performance of Robert edge detection at 200 iterations. (E) Performance of Prewitt detector at 200 iterations. (F) Performance of ACO detector at 200 iterations.

In our experimental processes, the image segmentation ability of ACO-ISA was investigated on iris areas of three types of typical textures, that is, normal texture, the radii Solaris, and the hyperpigmentation textures. After the processing of ACO-IRIS and ACO-TRA (texture representation ability, the alignment of each region to 3×3 sub-zones took place followed by the calculation of histogram for each sub-zone. The comparison of two regions was made by taking the mean of histogram distances for all corresponding pairs of sub-zones. A sum of 900 sample regions was involved for each kind of the texture. The average distances (interclass distance) are demonstrated in Table 2.1. It is obvious from the results that different texture regions are well-distinguished for the ACO segmentation technique compared to the results obtained from other techniques. Additionally, we also designed a self-organized feature map (SOFM) neural network a tool for texture recognition, and around 95.1% accurate recognition rate was attained. This unveiled that our proposed technique has the potential to be useful in automatic disease diagnosis systems using different types of images (iris, MRI, etc.).

Table 2.1. Interclass Spaces Between Different Iris Regions

Interclass Distance Radii Solaris Hyperpigmentation Normal Region
Hyperpigmentation 13.67 – 84.53
Normal region 97.95 84.47 –
Radii Solaris – 13.62 97.87

To scrutinize the performance of the ACO algorithm, real iris and brain MRI images were used in the experiments. The output results obtained from traditional edge detection techniques (Figs. 2.5B–F and 2.6B–E) exhibit the results with 200 iterations and the parameters were adjusted to β=3.5, η=0.08, v=0.017, p=1.3; while many ants were 10% of all the pixels. Higher values of v and smaller values of β lead to higher probability of probing new paths. Therefore for the deposition of further pheromones in a digital habitat, the number of iterations must be increased. Experimental results for the ACO-based technique showed promising results. The process times of ACO has been improved (shortened) compared to the processing times of ACO employed in former research [41]. However, ACO still requires more time than conventional edge detection techniques, as demonstrated in Table 2.2.

Table 2.2. Time Passed (in Seconds) by Traditional Edge Detection Techniques as Compared to ACO

Image Type Canny Robert Sobel Prewitt Artificial Bee ACO
Brain MRI 0.285 0.868 0.084 0.734 0.688 114.75
IRIS 2.67 0.085 0.078 0.067 0.078 60.08

Our current experiment involves the experimentation of five different images (five each of iris and brain MRI images). Figs. 2.5A and 2.6A display the original images while Figs. 2.5B–F and 2.6B–E show the results obtained from conventional algorithms (Canny, Robert, Prewitt, Sobel, etc.). Both of the image sets processed with the ACO technique exhibited superior performance compared to the images processed with conventional algorithms regarding efficiency and effectiveness. To validate the effectiveness of ACO results, experiments were conducted on various other (real and synthetic) images. The segmentation (using images other than iris and brain) results were filtered by the Robert, Canny, Sobel, Prewitt, and ACO edge detectors using the Java framework in the MATLAB toolbox. Number of ants in digital habitat were 10% of the total number of pixels and essential parameters were δ=0.03, β=3.5, v=0.017, p=1.3, λ=0.5, and f i =0.09. Tables 2.3 and 2.4 depict the difference in performance for various edge detectors.

Table 2.3. Edge Operator (Detector) Evaluation Table for Brain MRI Image

Sobel Prewitt Robert Artificial Bee Canny ACO
RMSE 5.5760 5.5826 5.9294 4.6394 4.6394 3.3519
SNR 43.0944 42.8559 29.2034 68.2989 68.2989 84.3721
PSNR 33.2047 33.1944 32.6719 34.8026 34.8026 37.6483

Table 2.4. Edge Operator (Detector) Evaluation Table for IRIS Image

Sobel Prewitt Robert Artificial Bee Canny ACO
RMSE 2.2728 2.2774 2.2940 2.1175 2.1175 1.8119
SNR 43.8058 43.4136 42.2534 55.4086 55.4086 67.5980
PSNR 41.0000 40.9827 40.9195 41.6154 41.6154 43.7026

Read full chapter

URL:

https://www.sciencedirect.com/science/article/pii/B9780128155530000021

Iris feature extraction using three-level Haar wavelet transform and modified local binary pattern

Prajoy Podder , ... Joarder Kamruzzaman , in Applications of Computational Intelligence in Multi-Disciplinary Research, 2022

1.4 Iris normalization

This section describes the iris normalization step. The size of different acquired iris images will vary because of the variation in the distance from the camera, angle of image capturing, illumination level, etc. For the purpose of extracting image features, the iris image is to be segmented and the resultant segments must not be sensitive to the orientation, size, and position of the patterns. For this, after segmentation, the resultant element is transformed to Cartesian. In other words, the circular iris image is transformed into a fixed dimension.

Fig. 1.2 illustrates the normalization of iris images from three datasets. For each of the datasets, the one original input image is shown, followed by its inner and outer boundary detection, and then its segmented version, and finally its normalized version. Fig. 1.2A describes Daugman's rubber sheet model for iris recognition. Three original images from three datasets are shown in Fig. 1.2B, F, and J. First of all, Fig. 1.2B is one original image from the CASIA-Iris-V4 dataset [33]. For the iris image in Fig. 1.2B–E represent the corresponding inner and outer boundaries, the segmented version, and the normalized version, respectively. Secondly, Fig. 1.2F is one original image from the CASIA-Iris-V1 dataset [34]. For the iris image in Fig. 1.2F–I represent the corresponding inner and outer boundaries, the segmented version, and the normalized version, respectively. Thirdly, Fig. 1.2J is one original image from the MMU iris database [35], and Fig. 1.2K–M represent the corresponding inner and outer boundaries, the segmented version, and the normalized version, respectively.

Figure 1.2. Illustrations of (A) Daugman's rubber sheet model; (B, F, J) original input images; (C, G, K) images with inner and outer boundary detection; (D, H, L) segmented iris regions, and (E, I, M) iris images after normalization.

Read full chapter

URL:

https://www.sciencedirect.com/science/article/pii/B9780128239780000058

Biometric Market Forecasts

Mark Lockie , in The Biometric Industry Report (Second Edition), 2002

3.3.6.1 Technical Developments

As a relatively new commercial entry to the biometric market, there are still plenty of technical developments being made in the iris recognition sector.

Iris recognition systems can operate in either an automatic capture mode or they can be used to capture the iris image manually. The automatic capture systems are easier to use (although more expensive), as there is no need for the user to assist with the aligning and focusing of the camera. Clearly a manual system is more difficult, as the user must move the sensor (or themselves) back and forth in order for their iris image to be captured.

Designing better image capture systems is a trend that is evident at present. One of the most recent announcements was by OKI Electric Industry (in association with Matsushita Communication Industrial Co). It launched a system (the Irispass-WG) which can automatically detect the position of a person's eyes and then scan the iris and the pupil. Panasonic is another supplier to have launched a device – the BM-ET500 Series access control system – which essentially operates in the same way by capturing the iris image in an automatic, user-friendly way.

Another market trend has been the launch of small, less expensive iris recognition devices. The best-known device is the Authenticam from Panasonic. The small size of the device and its compatibility with a range of middleware products, such as I/O Software's SecureSuite, makes it ideally suited for computer security applications. It can also be used as a Web camera. Another small IT device is the Irispass-h from OKI Electric Industry.

As a small note of caution, spoofing attempts made on the Authenticam device proved successful in the first half of 2002, although the manufacturer is believed to have made moves to rectify the situation by adding more of the anti-spoofing measures found in more expensive systems.

Away from the hardware developments, work in the software field has been continuing. Of particular note is the development of a new technique for altering the format of an iris template. This may prove important in alleviating some of the privacy concerns surrounding the use of large biometric template databases – in particular function creep. The idea is that a template's format can be made specific to a particular application, making it unworkable in other applications. The process of changing the format does not change the efficiency of the matching process, so no degradation to performance occurs.

Read full chapter

URL:

https://www.sciencedirect.com/science/article/pii/B9781856173940500086

Unconstrained Data Acquisition Frameworks and Protocols

João C. Neves , ... Hugo Proença , in Human Recognition in Unconstrained Environments, 2017

1.4.3.4 Other Biometrics: Iris, Periocular, and Ear

Commercial iris recognition systems can identify subjects with extremely low error rates. However, they rely on highly restrictive capture volumes, reducing their workability in less constrained scenarios. In the last years, different works have attempted to relax the constraints of iris recognition systems by exploiting innovative strategies to increase both the capture volume and the stand-off distance, i.e., the distance between the front of the lens and the subject. Successful identification of humans using iris is greatly dependent on the quality of the iris image. To be considered of acceptable quality, the standards recommend a resolution of 200 pixels across the iris (ISO/IEC 2004), and an in-focus image. Also, sufficient near infra-red (IR) illumination should be ensured (more than 2 mW/cm 2) without harming human health (less than 10 mW/cm2 according to the international safety standard IEC-60852-1). The volume of space in front of the acquisition system where all these constraints are satisfied is denoted as the capture volume of the system. Considering all these constraints, the design of an acquisition framework capable of acquiring good quality iris images in unconstrained scenarios is extremely hard, particularly at large stand-off distances. This section reviews the most relevant works and acquisition protocols for iris and periocular recognition at-a-distance.

In general, two strategies can be used to image iris in less constrained scenarios: (i) the use of typical cameras and (ii) the use of magnification devices. In the former, the Iris-on-the-Move system is notorious for having significantly decreased the cooperation in image acquisition. Iris images are acquired on-the-move while subjects walk through a portal equipped with NIR illuminators. Another example of a widely used commercial device is the LG IrisAccess4000. Image acquisition is performed at-a-distance; however, the user has to be directed to an optimal position so that the system can acquire an in-focus iris image. The need for fine adjustment of the user position arises from the limited capture volume of the system.

Considering the reduced size of periocular region and iris, several approaches have exploited magnification devices, such as PTZ cameras, which permit extending the system stand-off distance while maintaining the necessary resolution for reliable iris recognition. Wheeler et al. [58] introduced a system to acquire iris at a resolution of 200 pixels from cooperative subjects at 1.5 m using a PTZ camera assisted by two wide view cameras. Dong et al. [70] also proposed a PTZ-based system, and due to a higher resolution of the camera they were capable of imaging iris at a distance of 3 m with more than 150 pixels. As an alternative, Yoon et al. [59] relied on a light stripe to determine 3D position, avoiding the use of an extra-wide camera. The eagle eye system [60] uses one wide-view camera and three close-view cameras for capturing the facial region and the two irises. This system uses multiple cameras with hierarchically-ordered field of view, a highly precise pan-tilt unit, and a long focal length zoom lens. It is one of a few example systems that can perform iris recognition at a large stand-off distance (3–6 m). Experimental tests show good acquisition quality for single stationary subjects of both face and irises. On the other hand, the average acquisition time is 6.1 s which does not match with the requirements of real-time processing in non-cooperative scenarios.

Regarding periocular recognition at-a-distance, few works have been developed. In general, the periocular region is significantly less dependent on face distortions (i.e., neutral expression, smiling expression, closed eyes, and facial occlusions) than the whole face for recognition across all kinds of unconstrained scenarios. The work by Juefei-Xu and Savvides [62] is considered the only notable proposal to perform periocular recognition in highly unconstrained environments. The authors utilized the 3D generic elastic models (GEMs) [71] to correct the off-angle pose to recognize non-cooperative subjects. To deal with illumination changes, they exploited a parallelized implementation of the anisotropic diffusion image preprocessing algorithm running on GPUs to achieve real-time processing time. In their experimental analysis, they reported a verification rate of 60.7% (in the presence of facial expression and occlusions) but, more notably, they attained a 16.9% performance boost over the full face approach. Notwithstanding the encouraging results achieved, the periocular region at-a-distance still represents an unexplored field of research. The same holds for ear recognition. Ear is another interesting small biometric that has been proved relatively stable and has drawn researchers' attention recently [72]. However, like other similar biometrics (e.g., iris and periocular), it is particularly hard to be managed in uncontrolled and non-cooperative environments. Currently, the recognition of human ears, with particular regard to challenges of at-a-distance scenarios, has not been faced yet, thus representing a promising and uncharted field of research which could reserve interesting opportunities and achievements in the recent future.

Read full chapter

URL:

https://www.sciencedirect.com/science/article/pii/B9780081007051000014

Securing the elderly in cyberspace with fingerprints

Guanglou Zheng , ... Craig Valli , in Assistive Technology for the Elderly, 2020

3.6 Discussions

Authentication is based on "something you are" (e.g., a biometric marker), "something you know" (e.g., a password), or "something you have" (e.g., a smartcard). The biometric trait of fingerprints has been successfully used in forensics and digital identity verification systems. There are several other biometric traits that can be used for securing the elderly as well, such as iris, face, and ECG signals [19]. This section compares fingerprint-based biometric systems with other popular biometric identification technologies such as face- and iris-based recognition.

The iris-based authentication utilizes features extracted from iris of the patient to verify his/her identity. Feature generation from iris images, as proposed by Li et al. [20], includes procedures such as background removal, iris image normalization, generation 1D signals, signal processing by using wavelet transformation, and iris feature vector creation. Features generated from an iris have a unique and stable pattern and thus can offer a high matching performance in the authentication process. However, users normally have more concerns when their eyes get scanned compared to the scanning of their fingertips, since eyes are sensitive organ of the human body.

Face recognition–based biometric authentication uses rich features extracted from a facial image to authenticate a user. The system could be built using 2D or 3D images. Facial features are generated from a facial image using the following steps in sequence: face detection, normalization, and feature extraction [21]. Compared to the fingerprint recognition system, the facial image capture is contactless and is perceived to be more convenient and nonintrusive. Nonetheless, compared to the facial recognition technology, the fingerprint recognition technology has a longer history and is more mature. Forensic examiners have successfully used fingerprints for criminal identification for more than a century, which proves the accuracy of fingerprint matching [6].

A multimodal biometric authentication system could be developed in the future which amalgamates multiple biometric traits, for example, fingerprints, iris, and face. This system could achieve a higher level of security, since multiple traits are examined in the authentication process. Compared to a system relying on a single biometric trait, this approach can provide more feature information thereby improving the accuracy levels of a biometric system [22].

Read full chapter

URL:

https://www.sciencedirect.com/science/article/pii/B9780128185469000038

Circle and Ellipse Detection

E.R. Davies , in Computer and Machine Vision (Fourth Edition), 2012

12.10.1 More Recent Developments

Much work has recently been carried out on iris detection using the HT. Jang et al. (2008) were particularly concerned with overlap of the iris region by the upper and lower eyelids, and used a parabolic version of the HT to accurately locate their boundaries, taking special care to limit the computational load. Li et al. (2010) used a circular HT to locate the iris and a RANSAC-like technique for locating the upper and lower eyelids, again using a parabolic model for the latter: their approach was designed to cope with very noisy iris images. Chen et al. (2010) used a circular HT to locate the iris and a straight line HT to locate up to two line segment approximations to the boundaries of each of the eyelids. Cauchie et al. (2008) produced a new version of the HT to accurately locate common circle centers from circular or partial circle segments, and demonstrated its value for iris location. Min and Park (2009) used a circular HT for iris detection, a parabolic HT for eyelid detection, and followed this with eyelash detection using thresholding.

Finally, we summarize the work carried out by Guo et al. (2009) to overcome the problems of dense sets of edges in textured regions. To reduce the impact of such edges, a measure of isotropic surround suppression was introduced: the resulting algorithm gave small weights to edges in texture regions and large weights to edges on strong and clear boundaries when accumulating votes in Hough space. The approach gave good results when locating straight lines in scenes containing man-made structures such as buildings.

Read full chapter

URL:

https://www.sciencedirect.com/science/article/pii/B9780123869081000124

Appendix

Mark Lockie , in The Biometric Industry Report (Second Edition), 2002

6.1.6.2 The Eye

Two parts of the eye can be measured by separate biometric systems; the iris and the retina. The patterns within the eye – whether they are the retinal vascular pattern or the iris – are as unique to each person as a fingerprint, if not more. Even a person's own two eyes are distinct from each other. As well as the distinctness of these patterns, they also remain stable over a person's life, only being affected by a few diseases. Because of their close links to the brain, the eyes are one of the first parts of the body to decay after death.

The idea that the blood vessels of the retina could be used to identify individuals was published in the New York State Journal of Medicine as long ago as September 1935 by Drs Carleton Simon and Isidore Goldstein. The uniqueness of the retinal pattern was supported in the 1950s by Dr Paul Towers in his investigations of identical twins. He claimed that the retinal vein pattern provided the most difference between twins.

The first retinal scanning units were launched in 1985 by EyeDentify, which quickly established a monopoly in this market. Most of the early units were sold to military and other establishments requiring the high security that these systems offer. The company claimed that its systems have never let in an impostor. Despite this monopoly EyeDentify seems to have hit troubled times. Compounding its problems have been the emergence of a few other companies promising much smaller, cheaper and more user-friendly retina-based systems (although these have yet to hit the market).

Irises are also distinct physical characteristics. Two American ophthalmologists, Dr Leonard Flom and Dr Aran Safir studied the properties of the iris noticing that they have a highly detailed and unique texture and remain stable over periods of many years. The iris pattern itself is made up of furrows, crypts, corona, filaments, pits, freckles, striations and rings.

The first iris scanning systems were launched commercially in 1995 by US company IriScan (now know as Iridian Technologies), which claims to own all rights to the technology.

6.1.6.2.1 Mode of Operation

With the EyeDentification system the user places one eye about two inches away from the device, then stares at a green dot target to bring the eye into focus. Light from a 0.07   W source (less than a refrigerator bulb) is reflected off the retina, collected by a rotating lens, before conversion to an analogue signal. The lens scans the retina until it creates a good template.

The newer retina-based systems are being designed to scan a person's retina at arm's length.

Indian's system uses a monochrome camera with standard video optics to capture the iris image. Early models required close eye positioning, but this has now improved. The iris pattern has 266 independent variables. It is captured and digitised into an IrisCode for storage as a 256 byte template. Smaller devices designed as peripherals for use with PCs have also been launched.

6.1.6.2.2 Requirements of an Eye Scanning System

Public acceptability

Although there were early fears by some end users that retinal scanning would injure their eyes, their worries do not seem to have been passed onto the iris scanning systems. In fact, early comments that iris scanning would be unlikely to take off have been proved to be untrue with the emergence of systems in banks, airports and PCs.

Work and environmental factors

The eye is a very stable part of the human body and is only affected by a few rare diseases. As with all biometrics, however, there are people that the technology will find difficult to accommodate.

Ease of use

Both systems require a degree of user co-operation, although the iris system is less demanding of the user and recent systems developed actually locate and verify the iris pattern automatically.

Circumvention

Because of its close link to the brain, the eye is one of the first parts of the body to decay after death. The vein pattern changes within just a few minutes. This makes it very secure against attack, however low-spec iris systems have been fooled via a photo of the iris – although Iridian is working to correct this phenomenon.

Performance

Eye scanning biometric verification devices are the most accurate of all systems and can work in identification mode on large databases.

Read full chapter

URL:

https://www.sciencedirect.com/science/article/pii/B9781856173940500116