Fesenko A. Methods of data processing for identification and authentication systems based on eye biometrics.

Thesis is devoted to applied scientific research task to developing data processing methods for biometric identification systems by eye biometrics. The analysis of the main data pre-processing methods and models for identification systems based on eye biometrics was carried out. Were highlighted their main disadvantages. Basic requirements for the developing information security systems were analyzed. For the first time was proposed and developed approach to identify users by eye biometrics in which used few special procedures for forming the iris code, that allow to decrease time for finding iris area on the image and reduce time of formation the iris code without losing quality of identification. Developed integral-differential method of localization iris images, which through integration made area iris localization and by differentiation allocated pupils and iris limits, which allows to reduce computational complexity while maintaining high accuracy of the method. Developed the method of encoding iris using modified DOG-filter that allows encoding one image pixel by one bit of information and reducing iris code size and as a result reducing size of Access control system database, which have time privilege compared with using Gabor filter. Developed method for identifying users without iris image sample, by iris binary code and normalized Hamming distance using statistical Neyman-Pearson criteria. This approach to solving the problem reduce Access control system database size, which in turn decreases the access time to it. This approach makes it possible to solve two classes of problems - iris code search in the database to identify and preliminary control access to objects that are under the protection.