说明：  实现一种结合颜色空间、变换及变形模板的自动唇部定位及唇轮廓提取、跟踪方法首先在空间建立肤色模型进行人脸检测、定位, 并由人脸几何特征进行唇部粗定位然后结合唇色模型进行变换使肤、唇色差别明显化, 提出根据亮度信息对变换结果预处理后用法进行图像分割, 经唇色模型进一步验证后实现唇部精定位再使用变形模板来进行嘴唇轮廓特征提取, 为增强内轮廓定位的鲁棒性, 对经亮度预处理和唇色模型验证得到的口腔区域边缘图进行曲线拟合来实现内轮廓定位最后, 将唇读图像序列中上一帧的唇部定位结果拓展后作为当前帧的预测区域再进行处理来实现唇动跟踪。
(To achieve a combination of color space, transform and deformable template automatic lip localization and lip contour extraction, tracking the establishment of the first color model in the space of face detection, location, geometric features of the face by the rough location and then combined with lip lip transformation so that the skin color model, lip color difference visible, presented the results according to the luminance information to transform the image after pretreatment use of segmentation, and further validated by the lip model to achieve precise positioning and then use the deformation of the lip template for lip contour extraction , to enhance the robustness of location within the outline, by the brightness of the pretreatment and the lip model validation by oral Quyu edge map to fit curve to achieve the final positioning within the contour, the lip-reading image sequence in the last frame of the lips After positioning results extend the forecast area as the current frame)