报告题目：GPU-Accelerated Locally Injective Shape Deformation

报告人/单位：陈仁杰 / 德国马克斯普朗克计算机研究所

时间：2017年10月25日9:45-11:00

地点：管理楼1218

报告摘要：

We present a highly efficient planar meshless shape deformation algorithm. Our method is based on an unconstrained minimization of isometric energies, and is guaranteed to produce C ∞  locally injective maps by operating within a reduced dimensional subspace of harmonic maps. We extend the harmonic subspace of  [Chen and Weber 2015] to support multiply-connected domains, and further provide a generalization of the bounded distortion theorem that appeared in that paper. Our harmonic map, as well as the gradient and the Hessian of our isometric energies possess closed-form expressions. A key result is a simple-and-fast analytic modification of the Hessian of the energy such that it is positive definite, which is crucial for the successful operation of a Newton solver. The method is straightforward to implement and is specifically designed to harness the processing power of modern graphics hardware. Our modified Newton iterations are shown to be extremely effective, leading to fast convergence after a handful of iterations, while each iteration is fast due to a combination of a number of factors, such as the smoothness and the low dimensionality of the subspace, the closed-form expressions for the differentials, and the avoidance of expensive strategies to ensure positive definiteness. The entire pipeline is carried out on the GPU, leading to deformations that are significantly faster to compute than the state-of-the-art.

报告人简介：

陈仁杰目前任德国马普计算机研究所(Max Planck Institute for Informatics)高级研究员。2010年于浙江大学获得应用数学专业博士学位。毕业后曾任职于以色列理工大学(Technion� Israel Institute of Technology)、美国北卡罗来纳大学教堂山分校(The University of North Carolina at Chapel Hill)。研究领域为计算机图形学，主要研究方向包括几何处理、几何建模、计算几何及裸眼3D显示器等。