厦门大学学报（自然科学版）2011，Vol.50Issue(3)：509-514,6.
三维双线性多分辨率形变模型研究
Research on 3D Bilinear Multidimensional Morphable Models
摘要
Abstract
Morphable model has widespread applications in computer vision and computer graphics. The current model construction method eirher depends on the unsrable optical flow algorithm or it requires lots of human- machine interactions. The model matching procedure based on stochastic gradient descent algorithm ofren gets stuck in the local optimum. In this paper. 3D bilinear multidimensional morphable models are proposed. Firstly, dense correspondences between prototypic faces are constructed by face segmentation based on key feature and the mesh resampling. And then the model matching problem is converted into two bilinear programs after tight convex relaxations. Finally, the bilinear programs are efficiently solvable by modern interior point methods. Experiments with synthetic data and real image data validate that the 3D bilinear multidimensional morphable models outperform the morphable models in optimality, model matching speed, convergence rate and robustness to noise and outliers.关键词
形变模型/稠密对应/三维人脸重建/双线性规划Key words
morphable model/ dense correspondence/ 3D face reconstruction/ bilinear programming分类
信息技术与安全科学引用本文复制引用
刘碧霞,李绍滋,郭锋,苏松志..三维双线性多分辨率形变模型研究[J].厦门大学学报（自然科学版）,2011,50(3):509-514,6.基金项目
国家自然科学基金项目(60873179) （60873179）
深圳市科技计划基础研究项目(JC200903180630A) （JC200903180630A）
高等学校博士学科点专项科研基金(20090121110032) （20090121110032）
