The Application of Using Mosaic and Maximum Spanning Trees in Automatic Human Iris Recognition System

• Main Authors: SniperShih, 施嘉南
• Other Authors: 蔡超人
• Format: Others
• Language: zh-TW
• Published: 2003
• Online Access: http://ndltd.ncl.edu.tw/handle/83010987584748763216

碩士 === 國立臺灣科技大學 === 電機工程系 === 91 === Iris is one of the features of living things. It is the doughnut-shaped area between pupil and sclera. Iris itself is very delicate and weak, but it is still protected by eyelids, cornea and anterior chamber. Thus iris or eyes is one of organs which are well protected. As pointed in literatures, as if iris is not affected by ageing, all do its features remain unchanged from one-year-old till death. Except possessing the above stabilities, there is another characteristic, uniqueness. From medical reports, although identical twins are indistinguishable in appearance, their iris features are quite different. Even the iris features of the two eyes from one person have uniqueness, not to mention features from different identities. The foundation of human eye iris recognition system is based on the stability and uniqueness if iris. It can be said as biometric recognition system applied to entrance restrain, ATM etc, or e-economic identify recognition. The reasons why iris features is applied to recognition system , expect the above two characteristics, the most of all is that it is not easy to be copied. As a result, human eye iris recognition system is a reliable and safe recognition system. John Daugman developed the first iris recognition system using 2D Gabor filter and artificial neural network in 1993. Wides exploited serious iris recognition products using Laplacian-Pyramid in 1994 and W.W. Boles using 1D Wavelet transform. In This thesis, iris features are analyzed in 249 bytes iris code using mosaic and Maximum Spanning Trees (MST) method, and realize auto human eye iris recognition system by voice effects. In the first place, techniques for processing iris image are introduced. For blurred iris images, contrast stretching, histogram equalization or homomorphic filtering can be used to intense the image contrast, Description of iris images is done by space filter of first derivative or second derivative to stress iris texture. Low pass filters such as mean filter or gaussian filter are used to eliminate noise signals. In the following, iris code of iris image is constructed by Maximum Spanning Trees. At the last, auto human eye iris recognition system is completed by relativities of iris codes.