Noise Detecting In Iris Recognition

With the importance of information security, iris recognition has become a hot research topic for applied mathematics, pattern recognition, image processing and information security.The iris recognition system is mainly composed of iris image acquisition, image quality assessment, the iris image preprocessing, feature extraction and recognition. Iris image preprocessing is critical among the four parts. The iris image preprocessing mainly includes noise detection, iris localization and iris images normalization. Among them, this paper focuses on the noise detection. Most iris feature extraction and recognition algorithms have high requirements on the quality of iris images. Good-quality iris images are better than Low-quality iris images, but in the practical application, it is difficult to ensue the iris image quality. Most of the iris region of the iris image is affected by the eyelids and eyelashes.On the basis of introducing lots of noise detecting algorithms, this paper puts forward an eyelid location and eyelash detection algorithm to detect the eyelid border and remove eyelash noise, in order to improve the speed and efficiency of iris recognition. The innovative work in this article is as follows:In eyelid location, according to some knowledge of digital image processing such as Gabor filtering, morphology and gradient and so on, extract the binary image of eyelid boundary and utilize the modified parabolic to fit the eyelid boundary point, setting a criterion until to the optimal.After eyelid location, according to the parabolic shape, extract iris region from the iris image for eyelash detecting. Use Gabor filtering, Adaptive threshold and other relative knowledge of digital image processing to get the eyelash noise points.The iris noise mainly includes eyelid and eyelash. This article puts forward a new algorithm of eyelid location and eyelash detection to generate the iris noise template which can remove the impact of eyelid and eyelash.The algorithm put forward in this thesis has been programmed on Iris Database CASIA-IrisV3-Lamp via Matlab. The accuracy of lower eyelid is up to99.5%. Because of serious eyelid and eyelashes occlusion on CASIA-IrisV3-Lamp, the accuracy of upper eyelid is relatively lower. This is95.2%. The accuracy of eyelashes detection is98.6%.