An iris segmentation using harmony search algorithm and fast circle fitting with blob detection

Abstract

Pupil and iris segmentation based on ellipsis or circle recognition are sensitive to light reflections and reflected images. The method presented here is independent of size and shape and at the same time insensitive to light reflections and reflected mirror images. The pupil detected using the algorithm can be a reference point to further segmentation of the sclera of the eye as well as of the iris. The method is also effective when the pupil and iris are not positioned perpendicularly to the camera eye. The algorithm’s average segmentation accuracy for all tested databases was 96% when considering only noisy and distorted images whilst a result of 100% was achieved with unblurred and clear images. The proposed method can be quickly and simply reproduced with a combination of known image processing methods. The developed algorithm for detecting the eyelid boundaries is effective with noisy and poor quality images due to the use of edge approximation using the Harmony Search Algorithm. An optimized shape detection method was used to detect the pupil and its edges. A method based on the variation and the average was used to eliminate shadows and eyelashes. The proposed scheme was tested on the UBIRIS.v1 database, MMU.v1 database and MILES databases, providing high results and short segmentation time. Segmentation accuracy for UBIRIS.v1 was 98.14%, for MMU.v1 – 90% and for MILES – 99.8%.