基于主动形状模型的人脸面部特征定位方法研究

The study of the oriented method of man,s face feature on the basis of ASM

本文对传统的基于主动形状模型(ASM)进行人脸面部特征定位的方法进行了一系列的改进,第一我们在人脸检测的结果的基础上定位一些显著的特征点,利用这些特征点来进行模型的初始化,并在后续的搜索中进行位置的约束,以提高特征定位的精确性。根据已定位的显著特征点的位置,我们在对局部搜索结果重构之前对其位置进行了局部调整,这样就防止了搜索的结果偏离已定位的特征点情况的发生。第二,我们在局部灰度模型中加入了边缘约束,使边缘信息较强的点有更大的可能成为最佳候选点。

This article makes a series of improvements of the traditional oriented way of man,s face, which is based on ASM. Fist, we orientate some outstanding feature points on the basis of the results of face test, then make use of these feature points to initialize model and restrain the position in the follow-up search, in order to raise the veracity of feature orientation. We partly change their position before we search and restructure the part according to the position of outstanding feature points which have already orientated. So we prevent the searching results from departing form the feature points have orientated. Second, we add fringe restraint in the model of partial grey level, so that the points with stronger fringe information could have more possibility to become the perfect candidate points.