Font Size: a A A

Iris Preprocessing Method Study

Posted on:2012-03-05Degree:MasterType:Thesis
Country:ChinaCandidate:P YuFull Text:PDF
GTID:2208330332486764Subject:Access to information and detection technology
Abstract/Summary:PDF Full Text Request
With the rapid development of society, people require more safety of the information security. Iris recognition becomes a hot research area in recent years with its unique advantages. In the real-time iris recognition system, every frame from the camera must be evaluated correctly and quickly. This paper proposes a method, based on the optimal hyper plane, to evaluate the definition of every frame. In the process of iris segmentation, this paper proposes a circle pattern based on statistics. After the normalization of iris, this paper introduces a method, according to the phase congruency, to tackle noise, such as eyelid. The main task includes the followings:1. Segmentation of boundary between iris and pupil and evaluation of definition of the iris. First, locate one point in the pupil based on the gray level of pupil and spots in the pupil. Second, find the boundary points based on the gradient along the boundary and fit the pupil through Least Square. Third, this paper adopts two parameters: one is the sharpness of pupil boundary and the other is gradient engine of iris, to evaluate the definition of iris with the help of optimal hyper plane.2. Segmentation of boundary between iris and sclera. This paper proposes a circle pattern to locate the boundary accurately and quickly. One classifier is designed according to the distance between these candidate points and center of pupil. Finally, M-estimator is adopted for curve fitting of the iris.3. Tackle noise, such as eyelid, in the normalized iris image. This paper introduces one method, phase congruency, to tackle the noise.All the algorithms are simulated through Matlab7.0, and in VC system, average time consumption of evaluating every frame is less than 0.06 seconds, satisfying the need of real-time system. Rate of accurately segmentation of both inner and outer boundary is more than 99%.This system has a good performance: false accept rate is one in hundred thousands while the false reject rate is about 10%.
Keywords/Search Tags:iris recognition, definition evaluation, optimal hyper plane, iris segmentation, phase congruency
PDF Full Text Request
Related items