| With the rapid development of internet and computer information technology, information security has become more and more important. Thus there is a strong demand for reliable identification technology. With the development of the society, traditional identification techniques due to their limitations can not fully meet the requirements of people. In this environment, the biometric technology comes into people’s view. And because of its stability, convenience and reliability, the biometrics obtains a wide recognition. As a biometric technology, iris recognition is considered to be the most reliable owning to its higher stability, higher reliability and higher security. So the development and application prospects of iris recognition are very broad.Iris recognition system includes iris image capturing, image preprocessing, feature extraction and matching. In order to conduct iris recognition, it is first to acquire the iris images by the image acquisition system, and then iris images need to be treated in preprocessing. Generally speaking, the preprocessing step concludes iris location, eyelash and eyelid detection, normalization and image enhancement. After preprocessing, iris feature will be extracted in normalized images. The extracted iris feature information is expressed by some numbers and lastly we match the digitized iris information for identifying or recognition. Among these recognition steps, image preprocessing is a vital part because the result of preprocessing has a directly impact on the later step, feature extraction and matching, and determines the accuracy of the iris recognition system. This paper focuses on the exploration of eyelash and eyelid detection algorithm in preprocessing.In practice, acquired images in iris recognition always contain not only iris but also some undesirable parts, such as eyelid, eyelashes and so on. Such components will be considered meaningful and be coded as features, which will lead mismatch if they cannot be detected and removed. Therefore, a robust method is needed to stress such issue as much as possible.The exited eyelash detection methods always depend on the eyelid detection or the fixed threshold, they are often complex and not accurate. We propose an eyelash detection method based on the Expectation-maximization algorithm and Gaussian Mixture Model. It is supposed that an eye image’s intensity distribution can be represented by a mixture of Gaussian function. The eyelash region or the part composed of eyelashes and some other components in the image yielding to a Gaussian distribution. We need to find the parameter estimation of the Gaussian function of the eyelashes and then detect the eyelash region. Because of not relying on the detection of eyelid and not having to set the fixed threshold, our method solves the problems of complication and inaccuracy, so it is more practical.Then, this paper presents an eyelid detection method based on hybrid-edge detection. At first, it is need to conduct image filtering to eliminate the influence of eyelashes covering. Through dealing with the edge detection image after filtering, non-eyelid edge information is reduced and the eyelid fitting area is determined. We can locate the eyelid and avoid the determination of non-real eyelid quickly and accurately. Specifically, it is need to use order statistic filter to process iris images firstly. And then implement the edge detection with Canny operator of horizontal direction and obtain the first edge points from pupil center to upper boundary of the edge image to form the eyelid curve. Finally get the eyelid fitting region and carry out the curve fitting.In order to verify the effectiveness of the eyelash and eyelid detection algorithm proposed in this paper, we carried out the simulation experiments on the basis of iris image library CASIA1from Chinese Academy. Experimental results show that the proposed algorithm can detect the eyelashes and eyelid effectively, which has guiding significance to iris image quality evaluation and feature extraction. |
PDF Full Text Request