| Iris recognition as one of biological characteristic recognitions, springs up as a result of social and economical development, and shows great advantages. Iris image quality evaluation is an important step of iris recognition, because the image quality is crucial to recognition effect. Through analyzing some existing algorithms, and in allusion to the problems in iris image quality, a complete course of operations is proposed to evaluate iris image quality. The main tasks of this thesis include:First, iris localization is basal in the course. A new method based on the iris intensity value is proposed, in which we adopt edge points and curve fitting. Then, through three times of repeating calculation, the exact interior edge of the iris is found.Second, we evaluate the general quality of the image according to the characteristics of human eyes and the subjective evaluation merthods. It includes following steps: detect the brightness of whole image by analyzing the gray level histogram; estimate whether the iris is in the middle of image; calculate the distance between the eye and the camera by detecting the two reflective points.Third, we evaluate the quality of interesting regions. Through analyzing the gray value of iris texture, eyelash and eyelid, we calculate the usable part of iris; a new method is proposed to detect the eyelid below by edge operator; according to the characteristics of iris texture, the definition of iris is evaluated by edge energy function.Forth, image quality evaluation on real-time is realized. Different thresholds are established in different stages, only the images that meet the need of the thresholds are able to enter the next evaluation stage. In this way, evaluation time is reduced. The quality and speed of selected images is guaranteed.All the algorithms are tested in Matlab7.0, and in VC system the time of evaluating each image is less than 0.05s. It's proved to be valid to select qualified images and improve the iris recognition performance. |
PDF Full Text Request