A new hierarchical 3D facial correspondence method based on regional deformation and multi-constraint(HCRDM)is proposed to deal with the problem how to establish physiological consistent correspondence between 3D face samples according to morphological geometric characteristics in craniofacial statistical reconstruction. A set of regional geometrical template of facial organs is defined based on physiological structural landmarks so that a semi-automatic partition of face can be achieved. Then
each region in the sample is deformed based on radial basis function(RBF)according to the strict correspondence of landmarks. Voxel models are built for those approximately coincided corresponding regions
the current under-corresponding vertex(CUV)with largest curvature in the template is successively selected and a candidate set of corresponding vertexes on sample face is determined by the local geometrical constraint of CUV
which is generated from relative positions between those corresponded template vertexes and CUV. Then
the optimal corresponding vertex on the sample can be determined according to the weighted distances of local geometric features between CUV and each candidate vertex. The correspondences of border regions are finally established by the geometrical constraint and the distance-constraint of differential characteristics. Experimental results show that the accuracy of 3D facial correspondence established by HCRDM is 10% higher than that of existing algorithms.
关键词
Keywords
references
BESL P J, MCKAY N D. A method for registration of 3D shapes [J]. IEEE Transactions on PAMI, 1992, 14(2): 239-256.
SALVI J, MATABOSCH C, FOFI D, et al. A review of recent range image registration methods with accuracy evaluation [J].Image and Vision Computing, 2007, 25(5): 578-596.
BRETT A D, HILL A, TAYLOR C J. A method of 3D surface correspondence and interpolation for merging shape examples [J].Image and Vision Computing, 1999, 17(8): 635-642.
DAVIES R H, TWINING C J, TAYLOR C. Groupwise surface correspondence by optimizat-ion: representation and regularization [J].Medical Image Analysis, 2008, 12(6):787-796.
MIAN A S, BENNAMOUN M, OWENS R A, et al. Automatic correspondence for 3D modeling: an extensive review [J].International Journal of Shape Modeling, 2005, 11(2): 253-292.
HU Yongli, YIN Baocai, GU Chunliang, et al. 3D face reconstruction based on the improved morphable model [J].Chinese Journal of Computer, 2005, 28(10): 1671-1679.
FENG Jun, IP H H S, LAI Lap Yi, et al. Robust point correspondence matching and similarity measuring for 3D models by relative angle-context distributions [J]. Image and Vision Computing, 2008, 26(6): 761-775.
李康. 基于颅骨的人脸建模技术研究及在法医面貌复原中的应用[D]. 西安:西北大学, 2006.
张彦飞. 基于分区的颅面统计重建算法研究与实现[D]. 西安: 西北大学, 2010.
MA Hongjing, FENG Jun. Three-dimensional facial feature points matching based on a combined support vector machine [C]∥Proceedings of The First International Conference on Internet Multimedia Computing and Service. New York, NY, USA: ACM, 2009: 153-158.