الفهرس | Only 14 pages are availabe for public view |
Abstract With the development of mathematics along decades, the branch of quaternion mathematics has evolved as a solution to several complex problems. Quaternion mathematics are based on complex numbers that comprise a real part and three complex parts. Quaternion mathematics were introduced to solve the rotation problems of dynamic systems. This trend has come soon to all branches of science. It found applications in signal and image processing such as adaptive filtering, image encryption, and 3D graphics. In this thesis, the quaternions are used to represent color images acquired as biometrics. The three complex parts of the quaternion used in our work represent the color image components, while the real part is kept to zero. Quaternion mathematical operations are developed in this thesis to generate intentionally distorted versions of the biometrics to be used for identification instead of the original biometrics. The intended distortion effect is similar to the encryption effect as it masks the details of the original biometric images. This trend of using alternatives to original biometrics is known as cancellable biometrics. Its major objective is to save the original biometrics from being hacked if they are used in biometric databases. Three algorithms based on quaternion mathematics are presented in this thesis to generate cancellable biometric templates. The first one depends on certain quaternion multiplication processes between the original color biometric and an auxiliary image to induce the required degree of distortion in biometric templates. The second one depends on the utilization of FRFT with quaternion mathematics. The third one depends on the utilization of rotation of quaternions to get the required distortion level. Both algorithms are tested in the cancellable biometrics recognition framework using Equal Error Rate (EER) and Area under Receiver Operating characteristic (AROC) curve to validate the proposed cancellable biometrics scenarios. Simulation results have proved EER values close to zero and AROC values close to one. Hence, the proposed algorithms can be considered as good candidates for cancellable biometric security systems. |