【作者】 高志升；

【Author】 GAO Zhisheng(School of mathematic and computer,Xihua university,610039,ChengDu)

【机构】 西华大学数学与计算机学院；

【摘要】 一种彩色图像人脸检测新方法被提出。该方法首先对人脸彩色图像进行 PCA 变换以得到主分量 P,并结合图像亮度分量Ⅰ构建 PⅠ图。在 PⅠ图中使用经验掩模进行图像肤色的自动检测,并使用区域生长法得到肤色区域。再结合人脸的几何特征过滤噪声而得到候选人脸区域,并在候选人脸区域内结合眼睛和嘴的色度特征构造相应模板。最后使用该模板验证候选人脸区域并定位人脸特征点。实验表明该方法不仅具有能适应复杂背景的人脸检测的优点,而且检测率高,漏检率低。更多还原

【Abstract】 A novel method for face detect in color images is proposed.First the PI diagram is created by combining the prinincipal vector of image’s RGB vectors and illumination intensity.vector I.A mask is used in Pi diagram to detect the skin color automatically,then a growth algorithm is applied to get full skill color area.After getting ride rid of the noises by using the geometrical characters of facial in the skin area,candidate face area can be obtained.Eye map and mouth map can also be obtained by integrating the corresponding chrominance components of eyes and mouth area.At last the face can be validated by combining the candidate face area,eye map and mouth map.Experiment shows that not only can it be used to detect face in the more complex environment,but also has high detection rate and low miss detection rate.更多还原