论文部分内容阅读
Face recognition based on few training samples is a challenging task. In daily applications, sufficient training samples may not be obtained and most of the gained training samples are in various illuminations and poses. Non-sufficient training samples co