Iris Segmentation using Gradient Magnitude and Fourier Descriptor for Multimodal Biometric Authentication System
Defiana Sulaeman, Anto Satriyo Nugroho, Maulahikmah Galinium · Journal of ICT Research and Applications · 2016
Perfectly segmenting the area of the iris is one of the most important steps in iris recognition. There are several problematic areas that affect the accuracy of the iris segmentation step, such as eyelids, eyelashes, glasses, pupil (due to less accurate iris segmentation), motion blur, and lighting and specular reflections. To solve these problems, gradient magnitude and Fourier descriptor are employed to do iris segmentation in the proposed Multimodal Biometric Authentication System (MBAS). This approach showed quite promising results, i.e. an accuracy rate of 97%. The result of the iris recognition system was combined with the result of an open-source fingerprint recognition system to develop a multimodal biometrics authentication system. The results of the fusion between iris and fingerprint authentication were 99% accurate. Data from Multimedia Malaysia University (MMUI) and our own prepared database, the SGU-MB-1 dataset, were used to test the accuracy of the proposed system.