Iris recognition, as a new biometric technology of high reliability, is now seeing increasing applications in many aspects of the society such as access control, law enforcement, network security, etc. And many systems based on iris recognition are utilized at airports or banks to obtain accurate user authentication. Iris image preprocessing, which includes iris detection, iris segmentation, iris image quality evaluation, normalization and liveness detection, is an important component of iris recognition systems. However, work remains to be done to improve preprocessing. Furthermore, synthesis of iris images enables us to build a large database of iris images and thus, provides a platform for evaluation and comparison of iris recognition algorithms. Therefore, preprocessing and synthesis of iris images are focused in this thesis. The main contributions of our work reported in this thesis are as follows. 1.Two iris detection methods are proposed. The methods make use of the structure information of irises and give answers to whether there is an iris in an image and where it is. 2.Two novel fast methods for iris segmentation are described. The first method makes full use of special properties of circles and takes into account texture properties of different iris regions. Another method uses key points for iris segmentation. 3.Problems such as quality assessment of iris images, liveness detection and normalization are addressed. Quality assessment combines general and specific measures to meet the requirement of international standards. Then, methods for liveness detection are described. Finally, a simple iris deformation model is given to improve the accuracy of iris recognition algorithms. 4.Methods for the synthesis of iris images are proposed and a large database is built. To the best of our knowledge, it is the first large database of synthesized iris images. 5.An iris recognition algorithm based on local extreme points is proposed. The method uses local extreme points to represent 2D surfaces and obtains promising performance. Based on the above methods, we have developed an integrated iris recognition system and obtained satisfactory performance under real conditions.
修改评论