A Study On The Pre-Processing Algorithms For Iris Recognition

Iris recognition is a kind of novel biometrics which was developed from 1990s and it has attracted more and more attention because of its high accuracy. Thanks to its complex structures and abundant features, the iris has many desirable properties, such as unique, stable, non-intrusive, etc. Here, the property of non-intrusive is bound to be the trend of biometrics development.A typical iris recognition system has three main parts: the image capturing, the image preprocessing and the feature extraction and matching part. The image preprocessing part also has three parts: iris location, de-noising and normalizing, in order to dividing or withdrawing the iris region image in the human eye image, preparing for the feature extraction and matching.In this thesis, after the current research situation of iris recognition methods was comprehensively analyzed, an algorithm of image preprocessing is researched respectively, and finally a method of iris image preprocessing in the state of eyes opening unforcedly is proposed. The main works and researched results of this thesis are as follows.(1) This article has analyzed the questions of traditional localization method using round template. As a large amount of experiments show, unlike the outer edge of iris, the inner edge of iris is not an ideal circle, thus edges acquired though methods above may incorporate some information that does not belong to the iris or some iris texture information is cut, which may result in the pupil carry-over or the texture loss question that affect the iris feature extraction and recognition.(2) This paper proposes a real iris edge extraction method, in this method, the traditional localizing principle is not only broken, but also the process of noise detection is achieved by transiting the spatial domain into the frequency domain. The detailed steps are as follows: firstly, the light spot within the pupil is filled in the original image, then the image is unfolded into a rectangle and the circle detection is substituted by the point and line detection in the rectangle image to find the inner and outer edge, the normalization is also performed. This method not only solve the problem of pupil carry-over or the texture loss caused by that the inner edge is not an ideal circle but also sharply reduce the location time since it substitutes the point and line detection to the circle detection.(3) A noise detection method by using the 2D-Gabor filter is proposed. Firstly, the eyelid and eyelash are detected by using 2D-Gabor filter in the normalized iris images. Then the noise regions are marked. Finally, the interferential bits in the iris codes are eliminated to avoid the noise effect to the recognition as much as possible. This method can achieve pro-processing and enhance the accuracy of iris recognition.