Investigating edges and corners detection for biometric recognition : iris detection

Abstract

In today’s applications, the increasing need for high security is an essential reason behind searching for new methods in order to increase protection in the systems used. Biometrics presents a solution to increase authentication in the electronic devices like tablets and cell phones. The iris is one of the biometrics that could be used for such a target. The iris pattern uniqueness for each individual gives it an advantage over other ways of protection previously used like passwords. The work in this thesis will mainly focuses on how to recognize a person based on his iris pattern using a technique called Edge Based Corner Detector (EBCD). This technique consists of detecting the edges in the iris pattern and extracting the interest points called “corners”. The matching is based on a comparison of corners in two iris patterns based on the two characteristics specifying a corner which are the length ratio and the angle. In the following, there will be a detailed description about the EBCD technique and how it could be used in an iris recognition system. The steps followed goes from acquiring eye images, segmenting irises, applying the EBCD to reach the last step which is the matching and making the decision. To get a better performance, we added what we called preferred corners which will be explained through this report. These new features can replace the pigments and crypts that are difficult to detect. In the following, the overall algorithm and the performance of the two trials (before and after adding the preferred corners) are presented in this thesis.