Eyes reflect our desires, when we are looking to something that meanwhere our interest is.
Many applications threat this subject. Applications onHCI seem to be most interesting because of their daily use. On computers,smartphones and other technologies, gaze can offer comfortable manner tocontrol like for guiding by gaze directions (robotic control) 2, eyes pretendto become communication canal between interevent here, other applications likesecurity and control gain robustness by adding eyes detection and centre eyeslocalisation, in commercial sectors when analyse direction of potentialcustomer can help to sale product or improve its efficiency. All these recentapplications prove eyes localisation importance focussing on centre eyeslocalisation.Different methods of eyes localisation are summarized in 3 they canbe divided into two different categories according to sources used. (i) IR illumination images: it is about corneareflexion on this type of image which is very helpful for centre eyes location.(ii) light variations: the extreme variability of eyes appearance in realenvironment make eyes centre very complex challenge to achieve, differentapproaches are involved in this category, in this paper, we use afeature-based approach for robust and accurate eye centre localisation with amulti-stage scheme taking benefit of combining several methods, we makecontributions summarized in three points:(i) a fastapproach combining several methods. (ii) We introduceMSER regions for the first time in eyes localisation systems.
(iii) We applysimple technics like edge detection as a prepossessing step to reduce researchwindows. Furthermore, we evaluate robustess by using the very challenging BioIDdatabase. The obtained results are extensively compared with state of the artmethods for eye centre localisation.Precise Eye centre Locationmethods differs by computation, efficiency and more over precision degree.Latest one is defined by a metric which is considered nowadays to be the majorevaluation metric used, it was introduced by 4.
Three metricsrepresenting successively Eye region, Iris regions and Pupil (Centre eye)represent how much location nears truthy real centre eye. Yang and al work 5was one of the first impacting in Eye centre location using their specialfeatures in the frequency domain by proposing a novel GaborEyebased method whichmakes full use of the special gray distribution in the eye-and-brow region, andself adaptively. Others use Gabor filters like Hamouz et al6identifing tenfeature points on the face through gabor filters, S.Kimm7 use multiscaleGabor Feature vectors. 8 and 9 introduce isophote curvatures to infer thecentre of (semi)circular patterns and a novel centre voting mechanism, to perform very accurate eye centrelocation and tracking.10 use isophote curvatures in fatigue detectionprocess, others