改进的主动形状模型方法在人脸特征点定位中的应用  

Application of an Improved Active Shape Model Method in Face Features Location

作　　者：戈新良[1] 杨杰[1] 张田昊[1] 杜春华[1] 

机构地区：[1]上海交通大学图像处理与模式识别研究所

出　　处：《上海交通大学学报》2007年第8期1320-1323,1329,共5页Journal of Shanghai Jiaotong University

基　　金：上海市科委资助项目(03DZ14015)

摘　　要：针对主动形状模型(ASM)已成功应用于人脸特征点定位提出了两点改进方法.其一,首先用Adaboost方法在图像中检测到人脸区域,然后在人脸区域中检测到瞳孔的位置,为ASM中的点分布模型粗略地定位好初始位置;其二,将原始ASM方法中的关键点的1D纹理模型改进为基于核概率密度估计模型的2D纹理模型.改进的方法在SJTU人脸数据库中进行验证,结果表明,改进的ASM方法提高了特征点定位的精度.Active shape model （ASM） which has been applied successfully in the application of face features location was improved in two aspects. First,the face region is detected by Adaboost approach in the images, and then the position of pupils which provide roughly the initialization position for point distribution model of ASM is detected in face region. Second, 1D texture model of the original ASM is extended to 2D texture model which is based on the kernel probability density estimation model. This improved method was evaluated on SJTU face database, and the experimental results show that the improved ASM improves the accuracy of face features location.

关 键 词：主动形状模型 瞳孔检测 核概率密度估计 人脸检测 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]