【摘要】 在光照条件恶劣的情况下,基于主动表观模型(AAM)的人脸特征点搜索与定位算法的精度和鲁棒性会显著下降。本文提出采用全变分模型对图像进行光照预处理,通过对基于反射系数图的特征纹理建模来帮助特征点定位。通过实验对比,分析了本文算法在定位精度和鲁棒性上的提高。更多还原

【Abstract】 Under variable illumination,the robustness and accuracy of facial feature point detection algorithm based on Active Appearance Model are remarkably lost.To overcome this weakness,we preprocess the lighting images by Total Variation Model to get the reflectance images,and then we use reflectance images as the target of detection algorithm to locate all the facial feature points.Experiments show great improvements in detection results,by using our preprocessed lighting images algorithm.In this way,we can have a more accurate and more robust algorithm to detect facial feature points even under bad illumination.更多还原