Iris Segmentation in Visible Wavelength Images using Circular Gabor Filters and Optimization

Segmenting iris from images captured under the visible wavelength spectrum represents a challenging task. To address this problem, a new iris segmentation method based on the Circular Gabor Filter (CGF) and optimization is proposed. The CGF is utilized to localize the rough position of the pupil center. According to the localization result, few pixels of the pupil (iris) and non-pupil (non-iris) regions are set as initial pixels in the related region. Next, by optimization scheme, initial pixels are automatically propagated over the image until they cover the whole pupil (iris) and non-pupil (non-iris) regions. The originality of the proposed method includes developing a new method to localize the iris and eyelids boundaries simultaneously, and furthermore robust to reflections. Experimental results on the non-cooperative UBIRIS.v1 iris images reveal the effectiveness of the proposed method in comparison with some of state-of-the-art methods. The results include classification error rates of 2.08 and 3.05 % for session one and two of the UBIRIS.v1 database, respectively.