Red-eye is a phenomenon that the pupils of people appear unnaturally red when an image is captured using photoflash lamp. There are some algorithms existing for red-eyes detection, but all of them have less accuracy and cannot detect single red-eye. In this paper, a novel approach is proposed to automatically detect red eyes in photos, which is based on active appearance models, and a practicable automatic correction method is presented The goal of the algorithm is to detect all the red-eyes in photos without any human intervention. Finally, experimental results demonstrate the high accuracy and efficiency of the method.