As a reliable biometric for human identification, iris recognition has recently received increasing attention. There exist some successful commercial systems for iris recognition so far. However, it should be noted that current systems would normally work only if the iris images were taken with the cooperative users from a short distance. Hence, a problem naturally arises: how can we keep the iris recognition systems working normally if there are only lower resolution iris images? One effective solution is to improve resolution by means of super-resolution image enhancement. According to the characteristics of the iris texture, we pro- pose several methods for iris image resolution enhancement. This thesis is orga-nized as follows: ① Chapter 1 introduces the status quo of iris recognition in a distance and presents iris image enhancement technique in iris recognition systems. More- over, the significance of this research is introduced in this section. ② Chapter 2 provides a brief survey of research on super-resolution image en- hancement. It also reviews the state of the art of super-resolution image enhancement methods and evaluates their performance. Finally, some re- search challenges and future directions are discussed. ③ Chapter 3 describes a new iris image registration method, phase-correlation based registration. Compared with existing methods, this algorithm makes full use of iris characteristics and can compensate non-linear deformation of iris to some extent. ④ Chapter 4 presents a super-resolution algorithm for muti-frame iris images. This method can fuse spatial-temporal information of iris image sequence to enhance the resolution of iris images. ⑤ Chapter 5 describes a learning based iris image resolution enhancement method. In this method, the prior probability relation between different bands of iris frequency information is incorporated into image resolution enhancement algorithms to acquire the lost information for those seriously blurred iris images. In a word, in this thesis, we have made fruitful attempts to use super-resolution in our iris recognition system.
修改评论