An Eye Detector Based on Cues and Heuristics with a Good Accuracy/Complexity Trade-off

Eye detection plays a central role in automatic face detection systems and it is also important for face recognition and face tracking. Working on the normalized RGB color space (NRGB), we initially use a combination of corner identification, color and edges as cues for localizing eyes. The eyes localization process is aided by its application on modeled skin patches based on our previous work (Chen and Grecos, 2005). Subsequently, heuristics are applied on these localized cues for the final eye detectors. Experimental results show that our scheme is very fast in the AR and Champion databases, while retaining very high detection rates.