A pupil center detection algorithm based on eye color pixels differences
Catalin Ionescu, Cristian Foşalău, Daniel Petrişor, Cristian Zet · 2015
In this paper we present a new algorithm for pupil center localization. The face and eyes of the target subject are detected using the Cascade Classifier algorithm based on Haar-like features. A histogram equalization method is employed to increase image contrast. The core feature of our pupil center detection consists of an adaptive template matching procedure applied on the eye region image that integrates specific constraints related to location and color proprieties of areas surrounding the pupil in order to compute the precise pupil center. Experimental results demonstrate the efficiency of the algorithm by assessing the correctness of the pupil center detection with respect to a ground truth reference. This procedure will allow further development of a system that facilitates social interaction of impaired people by analyzing their eye movements and translating key zones of visual attention focus into sentences that alleviate their communication.