系列报告

 

交叉学科论坛

Interdisciplinary Forum

Images are commonly represented as bitmaps, and it is crucial to identify intrinsic geometric features of objects in such an unstructured format. Vectorization is a popular technique that converts raster images into a collection of parametric curves and surfaces, encoding the input's prominent features and yielding resolution-independent representations. In this talk, we propose variational principles for image vectorization along with effective algorithms based on the affine shortening flow and region merging, generalizing a steepest gradient descent for the reduced Mumford–Shah functional. We will also present recent applications in shape classification and historical glyph preservation.

何雨晨博士于2021年获佐治亚理工学院数学博士学位，2020年至2023年间先后在法国巴黎高等师范学校萨克雷分校、杜克大学及上海交通大学担任研究员，并于2023年加入香港城市大学数学系，任助理教授。其主要研究领域包括图像处理的变分方法、计算几何、数据驱动的反问题、深度学习理论及其在计算机视觉中的应用。相关研究成果发表于《SIAM Journal on Imaging Sciences》、《Journal of Computational Physics》、《Nature Communications》等国际权威学术期刊。