中文摘要：

复杂光照条件下的人脸特征定位对于实现非合作人脸重构识别具有十分重要的意义。提出了一种结合自适应区域增强和主动表观模型（AAM）的人脸特征定位方法。该方法采用基于多区域协作拟合的人脸轮廓提取方法提取复杂光照条件下的准确人脸区域，根据准确人脸区域纹理灰度分布将人脸白适应分割为光照良好和光照不足区域，并分别进行直方图均衡补偿，然后采用AAM算法进行特征定位以及边缘轮廓特征点校正。实验表明，提出的方法在复杂光照下平均点对点误差以及平均最近点线距误差方面均优于标准的AAM算法以及基于Retinex滤波的AAM算法，证明该方法在一定程度上补偿了光照不均匀对传统AAM算法的影响，提高了人脸特衙定位的准确度。

英文摘要：

Facial feature localization under complex illumination has a very important significance for realizing non- cooperative face reconstruction and recognition. In this paper a new facial feature localization method based on adap- tive region enhancement and active appearance model （AAM） is proposed to improve the facial feature localization result under varying illumination. A multi-region collaborative fitting based facial contour extraction method is used to extract accurate facial region under complex illumination. The face is divided into well illuminated region and in- sufficiently illuminated region adaptively according to the texture gray level distribution of the accurate facial region ; then adaptive region-based histogram equalization compensation is performed to minimize the illumination variation under different lighting conditions. The standard AAM algorithm is adopted to carry out facial feature localization and edge contour feature point calibration. The experimental results show that using the proposed method the average point to point error and the average nearest point line spacing error are superior to those using the standard AAM and Retinex filtering based AAM algorithms under complex illumination condition, which proves that the proposed algo- rithm can compensate the influence of uneven illumination on traditional AMM algorithm in a certain extent and im- proves the accuracy of facial feature localization.