A low complexity algorithm for eye detection and tracking in energy-constrained applications

This paper presents a low-complexity algorithm of non-intrusive real time eye-tracking by a single camera. The algorithm employs the feature-based approach, scanning a seven-segment rectangular filter over the input image to find the Between-The-Eyes pattern of human face and locating eyes within the pattern. Unlike related methods it neither uses computationally expensive cascades of Adaboost classifiers nor SVM template matching, while detecting eye with high accuracy. We implemented the algorithm and present the results of its experimental evaluation.