Lecture Series in Pattern Recognition
题 目（TITLE）：Image Segmentation Based on Level Sets of Probabilities
讲 座 人（SPEAKER）：Prof. Yizhou Yu，Department of Computer Science，University of Illinois at Urbana-Champaign
主 持 人 (CHAIR)：Prof. Xiaopeng Zhang
时 间 (TIME)：10:00AM, Mar 23 (Friday)，2012
地 点 (VENUE)：1115 Meeting Room
In this talk, I present a robust and accurate levelset based algorithm for interactive image segmentation. The level set method is clearly advantageous for image objects with a complex topology and fragmented appearance. Our method integrates discriminative classification models with the level set method to better avoid local minima. Our level set function approximates a posterior probabilistic mask of a foreground object. The evolution of its zero level set is driven by three force terms, region force, edge field force, and curvature force. These forces are based on a probabilistic classifier and an unsigned distance transform of salient edges. We further apply expectation-maximization to improve the performance of both the probabilistic classifier and the level set method over multiple passes. Experiments and comparisons demonstrate the superior performance of our method. At the end of my talk, I will briefly introduce a data-driven image color and tone style transfer technique, which appeared in SIGGRAPH 2011.
Yizhou Yu is an associate professor in the Department of Computer Science at University of Illinois, Urbana-Champaign, and a visiting professor at the University of Hong Kong. He received a PhD degree in Computer Science from University of California at Berkeley in 2000. He is a recipient of best paper awards at ACM SIGGRAPH/Eurographics Symposium on Computer Animation, National Science Foundation CAREER Award and Microsoft Fellowship. He is currently on the editorial board of the Visual Computer, Computer Graphics Forum and International Journal of Software and Informatics. His current research interests include computational photography, data-driven animation, geometry processing and video analytics.