电子科技
電子科技
전자과기
IT AGE
2012年
7期
15-17
,共3页
三维人脸%人脸对齐%形变模型%稠密对齐
三維人臉%人臉對齊%形變模型%稠密對齊
삼유인검%인검대제%형변모형%주밀대제
3D face%face alignment%morphable model%dense alignment
针对ICP、NICP、流光法用于对齐三维人脸，建立形变模型，存在人脸对齐精度和准确性上的局限性问题。在标准形变算法的基础上，改进了局部匹配点搜索策略和形变能量函数，增加匹配点的准确性。并根据非刚性形变算法的思想，提出三维人脸迭代稠密对齐方法。在BJUT－3DFace Data人脸数据上实验证明，文中算法提高了人脸对齐的精度和准确性。
針對ICP、NICP、流光法用于對齊三維人臉，建立形變模型，存在人臉對齊精度和準確性上的跼限性問題。在標準形變算法的基礎上，改進瞭跼部匹配點搜索策略和形變能量函數，增加匹配點的準確性。併根據非剛性形變算法的思想，提齣三維人臉迭代稠密對齊方法。在BJUT－3DFace Data人臉數據上實驗證明，文中算法提高瞭人臉對齊的精度和準確性。
침대ICP、NICP、류광법용우대제삼유인검，건립형변모형，존재인검대제정도화준학성상적국한성문제。재표준형변산법적기출상，개진료국부필배점수색책략화형변능량함수，증가필배점적준학성。병근거비강성형변산법적사상，제출삼유인검질대주밀대제방법。재BJUT－3DFace Data인검수거상실험증명，문중산법제고료인검대제적정도화준학성。
Methods like ICP, NICP, optical flow method, etc. have been put forward to align 3D face in order to construct the morphable model. But they have limitations in face alignment accuracy and the accuracy of established correspondences. The strategy in searching local correspondences and energy function are modified based on standard 3 D surface deformation method to improve the accuracy of correspondence. Finally, we present the 3 D face iterative dense alignment method based on the framework of non-rigid algorithm. Experiments based on BJUT-3D Face Data show that alignment accuracy and accuracy of correspondences are improved by our method.
