The Study On The Methods Of Feature Extraction Of Iris Texture

Iris diagnosis is a method which can be used to judge the potentially problem andlesions of human internal organs by checking the iris texture feature. Iridology is not atreatment method, but to remind people of the condition of the body. It helps people torecognize the subhealthy condition of the body, which is significant to discover the diseaseinstantly.Iris fiber structure texture density reflects the ability of the body to resist disease, if it isthin,it represents the disease resistance ability is weak, on the contrary, the disease resistanceability is strong. This paper mainly studies the density of the iris fiber structure texture,combined with methods of feature extraction of texture, and design a set of methods ofautomatic extraction, then uses Tamura algorithm and GLCM(Gray-level Co-occurrenceMatric) algorithm to extract the feature values, which constitute the eigenvector. By usingthe SVM(Support Vector Machine), we achieve detecting the density of iris fiber structuretexture, realize the classification and diagnosis of iris fiber structure texture. First, weprocess the preprocessing of the iris, extract the annular part of iris and normalize it torectangular, and according to characters of this part, we process an algorithm to extract a partas research object. According to the characters of density and methods of texture featureextractions, we choose coarseness of Tamura algorithm and energy and entropy of GLCM torepresent the density. At last, we choose the appropriate parameters by experiments, and thenuse SVM to classify the density, to realize the diagnose of iris density.This paper uses the above methods to extract the iris fiber structure texture of200pictures of iris, then uses SVM classification, the accuracy of the tested results is96.5%.This result shows that using texture extraction methods to represent the density of iris fiberstructure has a good effect.