Hybrid method for human eye detection

Eye detection is required in many applications in human-computer interaction, which plays an important role in screen control, user recognition and auto-stereoscopic displays. Considering the defects of traditional methods of human-eye detection, an accurate human-eye-detection algorithm has been proposed. This paper proposes a novel technique combining the Adaboost algorithm and a hybrid matching method. First, facial part in the whole image is located with Adaboost algorithm; the human-eye area is positioned through the hybrid feature extraction method. In extraction process, edge density, chrominance, HSV and skin color cues are applied separately. Some of the regions are then removed by applying rules that are based on the general geometry and shape of eyes. The remaining connected regions obtained through these four cues are then combined in a systematic way to enhance the identification of the candidate regions for the eyes. The proposed eye-detection algorithm effectively reduces the eye-detection candidate area and improves the detection accuracy.