Automatic Pupil Segmentation Based on Circular Active Discs

Biometric iris recognition technology is one of the popular recognition technologies in biometric authentication systems. Biometric iris recognition has been observed to provide high identification efficiency relative to the other biometric characteristic. This high performance is possible only through accurately segmenting the iris region from the given ocular image. Segmentation of pupil is one of the primary steps in iris segmentation. We propose a pupil segmentation method based on a circular active disc. As a prototype, this technique uses the active discs. The initialization of the active disc is automated using normalized template matching technique. Initialization is followed by an affine transformation of the active disc which makes the active disc lock on to the pupil. The optimum outline around the pupil region is achieved by making use of the Gradient Descent algorithm which is supplemented by efficiently computing the partial derivatives by using the Green's theorem. The algorithm has been validated on two publicly available databases Casia-V3 and the IIT Delhi iris database spanning 4598 iris images in total. The corresponding Dice index scores 0.9463 and 0.9308, respectively for the databases show the robustness of the algorithm, and a high degree of agreement with the ground truth.