An Efficient Iris Segmentation Technique based on a Multiscale Approach

The use of biometric signatures, instead of tokens such as identification cards or computer passwords, continues to gain increasing attention as an efficient means of identification and verification of individuals for controlling access to secured areas, materials, or systems and a wide variety of biometrics has been considered over the years in support of these challenges. Iris recognition is especially attractive due to the stability of the iris texture patterns with age and health conditions. Iris image segmentation and localisation is a key step in iris recognition and plays an essential role the accuracy of matching. In this paper, we propose a new iris segmentation technique using a multiscale approach for edge detection, which is a fundamental issue in image analysis. Due to the presence of speckles, which can be modelled as a a strong multiplicative noise, edge detection for iris segmentation is very important and methods developed so far are generally applied in one single scale. In our proposed method, we introduce the concept of multiscale edge detection to improve iris segmentation. The technique is effecient for edge detetcion, greatly reduces the search space for the Hough transform and at the same time is robust to noise thus improving the overall performance. Linear Hough transform has been used for eyelids isolation, and an adaptive thresholding has been used for isolating eyelashes. Once the iris is segmented, a normalization step has been carried out by converting an iris image from cartesien into polar coordinates which are more suitable to deal with rotation and translation problems. Extensive experiments have been carried out and results obtained have shown an effectiveness of the proposed method which provides a high segmentation success of 99.6%.