Study On The Pre-processing For Human Iris Image

Today’s society has entered the information age, and the information securityrequirements has become higher and higher. The traditional identification methods such aspasswords, credit cards can not meet people’s needs. Therefore, the identification methodsbased on biometric technology develop rapidly and apply widely. As one of the biometricrecognition technology with development potential, iris identification has become a hot topicin the biometric identification field because of its advantages of uniqueness, stability,non-invasive and can be captured. Iris preprocessing, a key technology of iris recognition,includes locating inner-outer boundaries, detecting noise, dividing effective region andnormalization etc. If the iris preprocessing fail, the effective region may be extracted with thenoise interference information and directly effect the accuracy of recognition. So, the irispre-processing will be studied specially in this paper.Iris locating is the foundation of iris preprocessing. First, three major locating methods,as the Differential-Integral operator locating method, the Hough transform method and thealgorithm based on human eye’s structural characteristics, are analyzed with simulation. Butthese methods have a common disadvantage for fitting the iris inner-outer boundaries as acircle, so a new algorithm based on the regional properties and iterative searching is proposedfor locating the iris. And for eliminating the effective of spot in the pupil and improving theiris locating accuracy, an algorithm based on the average level of region gray is used indetecting the reflection noise. Then comparing with the common locating methods, themethod in this paper can achieve more locating accuracy and speed.After analyzing the previous eyelid and eyelash detecting methods, a new algorithmbased on morphology and judging the relation between eyelid and eyelash occlusions isproposed for eyelid detecting. When detect the occlusion, only need simply judge which partof the eyelid to detect, not blind to detect the upper and lower eyelids of noise in order toeffectively improve the efficiency of the algorithm. In addition, an algorithm based on localgray average is used for detecting eyelash occlusion and compared with the threshold methodand hysteretic threshold method, this method has little affective by the illumination of lightand lower undetected rate.Because eyelash is the major factor of blocking the iris region, an algorithm of eyelashocclusions detection based on endpoint identification is proposed in this paper. Then thenormalization method based rubber paper models is used with the results of endpoints identification and eyelid detecting, a pure effective iris region is extracted. In order to verifythe validity of the results of the preprocessing,2-D Gabor filters are used to extract and codeiris effective region feature. And finally by matching the encoded information with theHamming distance method, the recognition accuracy of the iris preprocessing is proved.