报告人：邹永魁

报告题目：A Riemannian shape interpolation model and its numerical algorithm

摘要：This paper introduces a general shape interpolation model using arbitrary degenerate Riemannian metrics, and specializes it to an as-rigid-as-possible shape interpolation model by selecting a particular metric. Our main modeling idea is to decompose the tangent bundle into a screen distribution and a radical distribution, first computing geodesics on the screen and then optimizing along the radical directions, thereby overcoming the difficulties caused by metric degeneracy. For a specific metric induced by the as-rigid-as-possible energy, this structure leads to a tractable shape interpolation model. We then construct an iterative numerical method, whose key feature is that all variables admit closed-form updates, resulting in each iteration having complexity linear in both the number of interpolated shapes and the number of vertices in the keyframes. This efficiency allows our model to handle extremely high-resolution shape interpolation. Numerical experiments demonstrate that the proposed method generates visually high-quality and achieves a speedup of one to three orders of magnitude over existing ARAP-based and learning-based approaches.

报告时间：2026年5月12日8:00-9:00

报告地点：正心楼303

报告人简介：邹永魁，吉林大学suncitygroup太阳新城教授，博士生导师。主要从事随机偏微分方程数值方法的研究，在J. Sci. Comput.，CICP，J. Nonlinear Sci., Nonlinearity, Z. Angew. Math. Phys. 等学术期刊上发表学术论文50余篇，主持完成和在研国自然面上项目6项。