Image Enhancement on Nonideal Iris Recognition.

摘要

Iris recognition algorithms have been seen increased development in the past years as an effective biometric. However, many of the existing solutions have difficulty processing nonideal iris images obtained from uncooperative environments (e.g. subject is unaware of capture). Image mosaicking techniques in computer vision can aggregate the information from multiple images, and create a composite final image containing more information than any of the individual images. In our research, we tried to use the mosaicking techniques to improve the partial and off angle iris recognition performance. The dissertation contains three parts: 1) preliminary partial iris image mosaicking and the evaluation; 2) gaze angle estimation for off angle iris recognition; 3) subregion selection and mosaicking for off angle iris recognition.