Iris Recognition for Personal Identification
|School||Xi'an University of Electronic Science and Technology|
|Course||Communication and Information System|
|Keywords||Iris recognition Iris segmentation Automatic scale selection Rotation compensate|
With the increasing requirements for security, biometrics based personal identification methods have received extensive attention. Recently, iris recognition is becoming an active topic in biometrics due to its high reliability for personal identification. A great deal of progress in iris recognition has been achieved in the past decade, but many efforts remain to be taken to further improve the performance of iris recognition systems. Aiming to develop an iris recognition softer ware system, we investigate some of the key issues related to iris recognition in this thesis, including iris image segmentation, iris feature extraction and matching as well as the application of iris recognition in copyright protection of the digital media. The main contributions of our work reported in this thesis are as follows:1. An efficient iris segmentation method is achieved. In this method the accuracy boundary of pupil is obtained by an active counter model technique. After computing wavelet transform of the intensity signal of the normalized iris images along horizon direction, the position of eyelid occlusion edge points can be obtained by evaluating the behavior of the wavelet modulus maxima over scales, and the eyelid occlusion regions can be determined by polynomial-fitting. The eyelashes are detected by 1-D Gabor filter.2. An experiment platform is built based on the proposed segmentation method, where different iris feature extraction methods based on Gabor filters are tested and an ideal of designing Gabor filters for optimal iris texture separation is presented. Meanwhile, the iris recognition algorithm proposed by Daugman is implemented approximately to test the iris image segmentation method proposed in this paper on the built experiment platform.3. The micro-elements of iris image are analyzed by various image base functions and it is determined that LOG function is the base function dominating iris image generation. Then the LOG base is used to detect iris minutiae features with multi-scales. Moreover, an iris recognition system using automatic scale selection algorithm for iris feature extraction is proposed by researching on the automatic scale selection scheme for multi-scale analysis and compared with the most of the existing iris recognition methods. The resulting conclusions will be helpful for further research.4. Several rotation invariant algorithms are proposed for iris feature extraction and matching. Firstly, we present an iris recognition algorithm which adopts Zernike’s moment invariants to extract iris moment-based rotation invariant features. Based on the review of the efficiency and applied ranges of this method by emulation experiments, we propose an iris matching algorithm using the phase-correlations of the iris local regions. This method follows three stages: extracting regions of interesting (ROI); computing the Band-limited Phase-Only Correlation functions and evaluating the quality of ROI by giving a qualitative description of the occlusion degree; calculating the matching score by combining the BLPOC function and the evaluating results. The POC function can compensate for the image rotation affection on the iris recognition and the quality evaluation of the ROI can depress the degree of the operator participation that is required in the iris recognition system.5. An efficient scheme is proposed for protecting the ownership of the digital image content by combinning the iris recognition with digital watermarking technique. The main ideal is as follows. Iris image preprocessing is conducted firstly, and instead of the feature codes achieved by information compressing, an iris region of interesting at gray level is selected directly as the watermark. Then the watermark image is imbedded into the 1D cepstrum domain of the host image. Iris image authentication is achieved by computing the Band-limited Phase-Only Correlation (BLPOC) functions.