Autori: N. Popescu-Bodorin

Editorial: G. Elefterakis, S.Hannam, E. Kalyva, A.Psychogios, South-East European Research Centre, Infusing Research and Knowledge in South-East Europe - Proceedings of the 4th Annual South-East European Doctoral Student Conference (DSC 2009), 4, p.471-479, 2009.

Rezumat:

This paper presents a new approach to iris segmentation. The basic idea is that since the pupilar boundary is nearly circular, there must be a relevant, nearly circular, pupil concentric iris segment, situated around the pupil, containing the muscles which control the pupil movements and playing the most important role in iris recognition. There are two main computational procedures used here in order to find and extract the circular ring approximating the actual iris segment: k-Means Quantization and Run Length Encoding.
In other classical iris segmentation procedures, like those within the Wildes’s and Daugman’s models, iris segmentation means fitting circular contours by solving three-dimensional optimization problems for finding a radius and two center coordinates via gradient ascent or by using the Hough transform or by iterating active contours. Here, each boundary is found mainly by defining and solving an one-dimensional optimization problem.
The proposed Circular Fuzzy Iris Segmentation (CFIS) procedure is designed to guarantee that similar segmentation results will be obtained for similar eye images, despite the fact that the degree of occlusion may vary from an image to another. It consists in two steps: pupil finding and limbic boundary approximation. Both of them are fuzzy approaches based on k-Means and Run Length Encoding. The result of CFIS procedure is a circular ring (concentric with the pupilar boundary) which approximates the actual iris segment. It is not the best approximation for one specific iris segment but it is a stable one, meaning that for a class of similar eye images representing the same eye, similar circular rings are obtained (despite the variations introduced by occlusions and illumination conditions). When two similar images of the same eye are compared, each detected circular iris ring is pointing to the same physical support, possibly occluded by eyelids, eyelashes, specular and lighting reflections.

Cuvinte cheie: segmentarea irisului // iris segmentation

URL: http://fmi.spiruharet.ro/bodorin/