Multispectral Iris Recognition Analysis: Techniques and Evaluation

Multispectral Iris Recognition Analysis: Techniques and Evaluation by Christopher K. Boyce Master of Science in Electrical Engineering West Virginia University Lawrence Hornak, Ph.D., Chair Arun Ross, Ph.D. (Co-chair) This thesis explores the benefits of using multispectral iris information acquired using a narrow-band multispectral imaging system. Commercial iris recognition systems typically sense the iridal reflection pertaining to the near-infrared (IR) range of the electromagnetic spectrum. While near-infrared imaging does give a very reasonable image of the iris texture, it only exploits a narrow band of spectral information. By incorporating other wavelength ranges (infrared, red, green, blue) in iris recognition systems, the reflectance and absorbance properties of the iris tissue can be exploited to enhance recognition performance. Furthermore, the impact of eye color on iris matching performance can be determined. In this work, a multispectral iris image acquisition system was assembled in order to procure data from human subjects. Multispectral images pertaining to 70 different eyes (35 subjects) were acquired using this setup. Three different iris localization algorithms were developed in order to isolate the iris information from the acquired images. While the first technique relied on the evidence presented by a single spectral channel (viz., near-infrared), the other two techniques exploited the information represented in multiple channels. Experimental results confirm the benefits of utilizing multiple channel information for iris segmentation. Next, an image enhancement technique using the CIE L*a*b* histogram equalization method was designed to improve the quality of the multispectral images. Further, a novel encoding method based on normalized pixel intensities was developed to represent the segmented iris images. The proposed encoding algorithm, when used in conjunction with the traditional texture-based scheme, was observed to result in very good matching performance. The work also explored the matching interoperability of iris images across multiple channels. This thesis clearly asserts the benefits of multispectral iris processing, and provides a foundation for further research in this topic.