【摘要】 利用灰度积分投影直接对人脸图像进行检测和眼睛定位是一种常用的算法,但是直接采用该算法会受到背景、特征等因素的影响,识别准确率较低。提出了一种基于最大类间方差阈值和区域膨胀相结合的检测与定位算法。该算法首先计算最大类间方差设置阈值,把灰度图像转换为二值图像并检测出人脸区域,然后通过对该人脸区域中的连通区域进行膨胀及连通性处理,精确定位眼睛坐标。实验表明,此算法可靠,具有较好的识别效果。更多还原

【Abstract】 It is a usual algorithm to use gray-level integral projection for human face detection and eye location,but directly application is affected by many factors such as background and noise,which causes low precision.In this paper,we propose a algorithm based on maximum variance between classes and area expansion.First,we set the threshold by calculating maximum variance between classes,and then transfer the grey-level image to binary image and detect the human area,finally,the connected domains of human face can be expanded and joined together,then we can get the eye location.The experiments show that this method is reliable and have a good precision.更多还原