视频序列中人脸眼嘴特征定位

Location of the feature of human eyes and mouth in video sequences

提出了一种基于数学形态学和分水岭算法的眼嘴特征定位算法 .根据人脸稳定的色度、梯度和亮度空间分布的信息 ,先利用数学形态学的运算对图像进行有效部分加强的预处理 ,后利用分水岭算法对加强图像进行分割 ,然后再定位特征 .实验证明该算法能有效定位正面人脸的眼嘴高级特征 ,对于轴线不与视平面平行的脸部和一些被阴影遮挡脸部也能确定其大致位置 .

A locating algorithm of human eyes and mouth features w as presented based on morphologic and watershed algorithms. According to stable character distributing of the face chroma, grads and luminance, morphologic algo rithm was used to enhance image, then watershed algorithm was used to segment th is image, and finally the eyes and mouth were located. Experiments showed that t he method was not only able to extract eyes and mouth features for right face ef ficiently, but also could find out their positions approximately for not right f ace and faces sheltered by shadow.