基于非刚性ICP的三维人脸稠密对应算法
|
摘要
三维人脸稠密对应是三维人脸分析研究的前提和基础。目前大多数的稠密对应技术是基于模板形变的方式,非刚性最近点迭代(iterative closest point, ICP)是应用最为广泛的一种,该算法通过逐步形变一个高分辨率的三维人脸模板来逼近目标人脸(扫描人脸数据)。但该类方法通过牺牲边缘精度来保持边缘区域的拓扑结构,以保证人脸之间的稠密对应。针对这一问题,提出了一种结合拓扑结构损失项的非刚性ICP算法,使得保持边缘区域拓扑结构的同时不会大幅度牺牲配准精度。实验结果表明,该算法比目前广泛使用的算法配准精度更高。
The dense correspondence of 3 D facial surface is the premise and foundation of 3 D face analysis. The majority of the techniques deform a template to fit the target, and the non-rigid iterative closest point(ICP) is the most widely used one, this method fits high-resolution 3 D facial template mesh to 3 D target face(scan face data) step by step. But this type of methods maintain the topology of the edge regions at the expense of edge precision, in order to ensure the dense correspondent between deformed faces. Aiming at this problem, a non-rigid iterative closed point algorithm combining topological structure loss was proposed, which keeps the topology of the facial edge regions without sacrificing the registration accuracy. Experiments show that our algorithm perform better than the state of art.
The dense correspondence of 3 D facial surface is the premise and foundation of 3 D face analysis. The majority of the techniques deform a template to fit the target, and the non-rigid iterative closest point(ICP) is the most widely used one, this method fits high-resolution 3 D facial template mesh to 3 D target face(scan face data) step by step. But this type of methods maintain the topology of the edge regions at the expense of edge precision, in order to ensure the dense correspondent between deformed faces. Aiming at this problem, a non-rigid iterative closed point algorithm combining topological structure loss was proposed, which keeps the topology of the facial edge regions without sacrificing the registration accuracy. Experiments show that our algorithm perform better than the state of art.
引文
[1] 项聪颖,周大可,杨欣.基于测地线环的三维人脸快速识别算法[J].电子测量技术,2018,41(4):81-86.
[2] ZHU S,LI C,CHANGE LOY C,et al.Face alignment by coarse-to-fine shape searching[C].Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition,2015:4998-5006.
[3] BOLKART T,WUHRER S.3D faces in motion:Fully automatic registration and statistical analysis[J].Computer Vision and Image Understanding,2015,131:100-115.
[4] ELSEBERG J,BORRMANN D,NüCHTER A.One billion points in the cloud-an octree for efficient processing of 3D laser scans[J].ISPRS Journal of Photogrammetry and Remote Sensing,2013,76:76-88.
[5] SALAZAR A,WUHRER S,SHU C,et al.Fully automatic expression-invariant face correspondence[J].Machine Vision and Applications,2014,25(4):859-879.
[6] 胡阳明,周大可,鹿乐,等.基于改进ASM的三维人脸自动对齐算法[J].计算机工程,2013,39(3):250-253.
[7] ZULQARNAIN G S,MIAN A,SHAFAIT F,et al.Dense 3D Face Correspondence[J].IEEE Trans Pattern Anal Mach Intell,2014,40(7):1584-1598.
[8] BLANZ V.A morphable model for the synthesis of 3D faces[C].Conference on Computer Graphics and Interactive Techniques,ACM Press/Addison-Wesley Publishing Co.,1999:187-194.
[9] PATEL A,SMITH W A P.3D morphable face models revisited[C].Computer Vision and Pattern Recognition,2009(CVPR 2009),IEEE,2009:1327-1334.
[10] COSKER D,KRUMHUBER E,HILTON A.A FACS valid 3D dynamic action unit database with applications to 3D dynamic morphable facial modeling[C].IEEE International Conference on Computer Vision (ICCV 2011),2011,156(1):2296-2303.
[11] ALLEN B,CURLESS B.The space of human body shapes:Reconstruction and parameterization from range scans[C].ACM,2003:587-594.
[12] AMBERG B,ROMDHANI S,VETTER T.Optimal step nonrigid ICP algorithms for surface registration[C].Computer Vision and Pattern Recognition IEEE Computer Society Conference,2007:1-8.
[13] LIANG L,WEI M,SZYMCZAK A,et al.Nonrigid iterative closest points for registration of 3D biomedical surfaces[J].Optics & Lasers in Engineering,2018,100:141-154.
[14] BOOTH J,ROUSSOS A,PONNIAH A,et al.Large scale 3D morphable models[J].International Journal of Computer Vision,2018,126(2-4):233-254.
[15] 严冬明,胡楷模,郭建伟,等.各向同性三角形重新网格化方法综述[J].计算机科学,2017,44(8):9-17.
[2] ZHU S,LI C,CHANGE LOY C,et al.Face alignment by coarse-to-fine shape searching[C].Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition,2015:4998-5006.
[3] BOLKART T,WUHRER S.3D faces in motion:Fully automatic registration and statistical analysis[J].Computer Vision and Image Understanding,2015,131:100-115.
[4] ELSEBERG J,BORRMANN D,NüCHTER A.One billion points in the cloud-an octree for efficient processing of 3D laser scans[J].ISPRS Journal of Photogrammetry and Remote Sensing,2013,76:76-88.
[5] SALAZAR A,WUHRER S,SHU C,et al.Fully automatic expression-invariant face correspondence[J].Machine Vision and Applications,2014,25(4):859-879.
[6] 胡阳明,周大可,鹿乐,等.基于改进ASM的三维人脸自动对齐算法[J].计算机工程,2013,39(3):250-253.
[7] ZULQARNAIN G S,MIAN A,SHAFAIT F,et al.Dense 3D Face Correspondence[J].IEEE Trans Pattern Anal Mach Intell,2014,40(7):1584-1598.
[8] BLANZ V.A morphable model for the synthesis of 3D faces[C].Conference on Computer Graphics and Interactive Techniques,ACM Press/Addison-Wesley Publishing Co.,1999:187-194.
[9] PATEL A,SMITH W A P.3D morphable face models revisited[C].Computer Vision and Pattern Recognition,2009(CVPR 2009),IEEE,2009:1327-1334.
[10] COSKER D,KRUMHUBER E,HILTON A.A FACS valid 3D dynamic action unit database with applications to 3D dynamic morphable facial modeling[C].IEEE International Conference on Computer Vision (ICCV 2011),2011,156(1):2296-2303.
[11] ALLEN B,CURLESS B.The space of human body shapes:Reconstruction and parameterization from range scans[C].ACM,2003:587-594.
[12] AMBERG B,ROMDHANI S,VETTER T.Optimal step nonrigid ICP algorithms for surface registration[C].Computer Vision and Pattern Recognition IEEE Computer Society Conference,2007:1-8.
[13] LIANG L,WEI M,SZYMCZAK A,et al.Nonrigid iterative closest points for registration of 3D biomedical surfaces[J].Optics & Lasers in Engineering,2018,100:141-154.
[14] BOOTH J,ROUSSOS A,PONNIAH A,et al.Large scale 3D morphable models[J].International Journal of Computer Vision,2018,126(2-4):233-254.
[15] 严冬明,胡楷模,郭建伟,等.各向同性三角形重新网格化方法综述[J].计算机科学,2017,44(8):9-17.