CAJ | 학술논문

人脸特征点的精确定位一直是人脸图像处理的重要研究内容，特征点定位精确与否直接影响后续工作结果的好坏。在基于反向组合AAM（Active Appearance Models）人脸特征点定位算法的基础上，提出结合特征点局部纹理模型来对AAM初始形状参数做最优化以及对AAM匹配模板升级的改进。改进的算法采用特征点局部纹理模型和AAM全局纹理模型结合的方法来最优化AAM初始形状参数，并在此前提下对AAM匹配模板进行升级，使其更接近待匹配图像的信息。在精确的匹配模板和优化的初始形状参数下，匹配的最终精度会得到提升。实验和理论证明，改进后的算法比传统反向组合AAM算法以及现有改进的PAAM（Progressive AAM）算法以及简单的结合ASM和AAM的改进算法都有更好的特征点定位精度。
인검특정점적정학정위일직시인검도상처리적중요연구내용，특정점정위정학여부직접영향후속공작결과적호배。재기우반향조합AAM（Active Appearance Models）인검특정점정위산법적기출상，제출결합특정점국부문리모형래대AAM초시형상삼수주최우화이급대AAM필배모판승급적개진。개진적산법채용특정점국부문리모형화AAM전국문리모형결합적방법래최우화AAM초시형상삼수，병재차전제하대AAM필배모판진행승급，사기경접근대필배도상적신식。재정학적필배모판화우화적초시형상삼수하，필배적최종정도회득도제승。실험화이론증명，개진후적산법비전통반향조합AAM산법이급현유개진적PAAM（Progressive AAM）산법이급간단적결합ASM화AAM적개진산법도유경호적특정점정위정도。
The precise localization of face feature points is always an important research contents in face image processing, the accuracy of the face feature localization has a directly affect of the result of the follow-up work. In order to achieve a high-precision of the face feature localization, the local texture models is introduced to optimize the initial parameter of Active Appearance Models(AAM)and the upgrade of the matching template of AAM after the study of AAM-reversed algorithm. To optimize the initial parameter, the local appearance model is combined with the global texture model of AAM, the matching template of AAM is updated to make sure that it’s closer to the real-time image. With the more precise template and the optimized parameter, the precise localization of face feature will be improved. It has been improved that algorithm has a better accuracy than the traditional AAM-reversed algorithm and the Progressive AAM(PAAM)algorithm through the theory and experimental.