关键词: 人眼探测, 概率主成分分析, 红眼效应, 主成分分析

Abstract: In a traditional stereoscopic display system, the user has to wear HMD or stereo glass,which makes the user uncomfortable and inconvenient. In contrast, an autostereoscopic display system can provide great enjoyment of stereo visualization without the drawbacks mentioned above.In order to render the stereo video into the left and right eyes of the user,who can move freely,we have to obtain the positions of the user’s eyes.As a result, eye detection is an important part in the autostereoscopic display system.In this paper,we propose a new method, based on the redeye effect, to detect the human eyes.The algorithm uses the difference between pupil illuminated by the on-axial and the off-axial infrared light,candidates of eyes can be segmented by subtract the two difference image.Principal Component Analysis and Probabilistic Principal Component Analysis are proposed for eye recognition.Experimental results show that the algorithm is very accurate and robust.

Key words: eye detection, PPCA, red-eye effect, PCA