Ivan A. Matveev

Iris Segmentation System Based on Approximate Feature Detection with Subsequent Refinements

A system of methods for iris region location and segmentation in frontal eye images is presented. Input data are images used in modern iris recognition systems, output contains coordinates of inner and outer iris borders and mask of visible iris region or decision that image does not contain iris of acceptable quality. System processing starts with approximate detection of eye center position followed by approximate detection of inner and outer borders of iris region. If one of the borders is not detected, an additional attempt to locate it is performed with the help of alternative methods. Finally precise borders are detected at last steps of processing by specially designed method.

3D surface reconstruction in automatic recognition system

Development of computer vision systems raises a problem of creating effective methods for restoring surface characteristics from their photographic images minimizing calculations while preserving reasonable accuracy of measurements. Here optimized algorithms of the surface reconstruction from stereo-images is proposed. The optimization is achieved by introducing some constraints on the conditions of image recording and by involving an a priory affirmation about objects in study. Effective algorithms have been developed on the basis of correlation method estimating corresponding points on images in stereo set and disparity map computation. Performance optimization is achieved by some modifications of correlation method including a pyramidal presentation of the images. The system uses a stereoscopic registration of face images and performs real-time image processing. The system was tested on a database of stereo- images.

Automatic stereoscopic system for person recognition

A biometric access control system based on identification of human face is presented. The system developed performs remote measurements of the necessary face features. Two different scenarios of the system behavior are implemented. The first one assumes the verification of personal data entered by visitor from console using keyboard or card reader. The system functions as an automatic checkpoint, that strictly controls access of different visitors. The other scenario makes it possible to identify visitors without any person identifier or pass. Only person biometrics are used to identify the visitor. The recognition system automatically finds necessary identification information preliminary stored in the database. Two laboratory models of recognition system were developed. The models are designed to use different information types and sources. In addition to stereoscopic images inputted to computer from cameras the models can use voice data and some person physical characteristics such as persons height, measured by imaging system.

Detecting precise iris boundaries by circular shortest path method

The problem of detecting precise pupil border in eye image given its initial circular approximation is addressed with circular shortest path method. Brightness gradient direction is employed to choose image pixels, which may belong to pupil boundary. Using initial approximate circles allows the method to work in a narrow ring, which contains only single pupil contour. Under these conditions the method allows to correctly handle almost all images used for iris recognition tasks and appears to be more precise than human expert in marking the pupil border. The method was tested with public domain iris databases, containing more than 80000 images totally. Experiments show that refinement of pupil border increases precision of iris recognition.

Location of pupil contour by Hough transform of connectivity components

A method for determining the pupil boundary in the image of eye is proposed. The method is based on image binarization followed by a search of the pupil as one of the connectivity components. The pupil boundary is determined as a part of boundary of the connectivity component. Hough transform is used for separating pupil in the case of its merging in one connectivity component with other objects, as well as to verify the likelihood of solution.

Method of multimodal biometric data analysis for optimal efficiency evaluation of recognition algorithms and systems

A primary consideration of this paper is to determine different factors influencing the reliability of performance evaluations of remote person recognition algorithms and systems. The authors suggest a method for determining and computing quantitative quality criteria of multimodal biometric data and consider the possibility of extrapolating test results to various practical applications. The functions of biometric data quality and biometric data artificiality that are introduced as a measure of proximity of the available biometric data to biometric data registered “naturally,” i.e., data of unaware and noncollaborative subjects, are under examination in this paper.

Detection of iris in images using brightness gradient projections

A method is proposed to detect human iris location and size in digital image given some point lying inside the pupil. The method is based on construction of histograms, or projections of local brightness gradients, and combinations of projections’ maxima being regarded as possible locations of pupil and iris borders. The method is notable for its low computational complexity and high tolerance to noise.

Optimization of informative components for 3D object recognition

Work presented suggests a combined informational space and decision rule for recognition of 3-D objects. The informational space consists of heterogeneous sets of features (i.e. belonging to different spaces), that are object images, images of certain object features and 3-D object surface representation. Decision rule for recognition in this combined space is proposed. The method was tested on a database of human face stereo-images and gave a significant improvement of reliability of automatic recognition system.

Locating the Visible Part of the Iris with a Texture Classifier with a Support Set

Person identification by the iris is one of the leading technologies in biometric identification. The visible region of the iris has the form of a ring enclosed between the pupil and the sclera partially occluded by eyelids, eyelashes, and flashes. An important problem is to find the non-occluded part, i.e., divide the pixels of the image into two classes: “iris” and “occlusions.” We propose an approach to solving this problem based on distinguishing a support set, i.e., a part of the ring which is free from occlusions with high probability, and subsequently finding all elements that have similar texture features. As the support set, based on experiments we have chosen a sector of the ring with minimal brightness excess. We divide the pixels with a classifier based on a multidimensional Gaussian trained on the support set. Local classification noises are partially removed by morphological postprocessing. Applying this algorithm to construct biometric templates improves recognition.

Influence of degrading factors on the optimal spatial and spectral features of biometric templates

Abstract Various distortions in the source data not only decrease the overall identification performance of a biometric system, but also alter the optimal parameters of a template creation method. In this work the influence of distortions to wavelength and spread parameters of the wavelets is presented. Three types of source data degrading factors are investigated: image blurring, image noise and iris segmentation errors. Two most popular methods of template creation, Gabor and Log-Gabor transforms are involved. CASIA and NDIRIS public domain databases are used for tests. It is shown that the optimum wavelength is strongly altered by image degradation whereas the optimal ratio of wavelength to spread, which defines filter shape, stays almost constant.