An Efficient Algorithm of Iris Segmentation for Mobile Devices

• Main Authors: TSOU,YU-HAO, 鄒裕浩
• Other Authors: TSAI,CHUNG-HSIEN
• Format: Others
• Language: zh-TW
• Published: 2018
• Online Access: http://ndltd.ncl.edu.tw/handle/z56vn3

碩士 === 國防大學理工學院 === 資訊工程碩士班 === 106 === The thesis proposed an efficient iris segmentation method of the iris images for mobile devices. Due to the complicated texture of the iris images and several critical factors from mobile devices using visible light to extract iris image, it will affect the iris segmentation results. These factors includes involuntary movements of the eye ball, low resolution and out of focus of the camera lens, environmental noise, and so on. The blurry segmented iris image of mobile devices could not be applied for further applications. In order to acquire iris images, the conventional iris image researches mainly utilized infrared light to extract and to segment iris images under controlled environment. For example, iris circle was detection proposed by John G. Daugman. However, the methods from these researches still need high computation cost and time. The thesis based on different image preprocessing filters to extract adaptive feature points to assist to transform imperfect orthographic iris images by rotation matrix. The proposed method can find to segment correct iris position efficiently. The proposed used the UBIRIS.v2 dataset applied to conduct a series of experiments in this thesis. Therefore, the contained noise of iris images in the database can be regarded as the mentioned factors of iris image segmentation problems of mobile devices. After verified and analyzed the experimental results, among different image preprocessing filters, the outcome can be found that iris images applied Median Filter and Histogram Equalization filter can further integrate orthogonal matrix to segment iris image most accurately and efficiently. In the future work, the proposed method can be further combined with machine learning and cloud computing techniques to expand personalized computing services.