|
楼主 |
发表于 2004-6-25 20:22
|
显示全部楼层
北京大学科学与工程计算系
日期: 06/23/2004 11:05 上午
主题: 学术报告预告
应用数学学术报告之十九
主 题:Image processing using PDE tools and Shape identification using level set methods
主讲人:Prof.Xue-Cheng Tai
( Department of Mathematics University of Bergen, Norway.)
时 间: 6月25日(星期五)16:00-17:00
地 点: 北京大学理科一号楼1560S
Tea Time:15:30 -16:00
地 点: 北京大学理科一号楼1384E
主 办: 北京大学数学研究所
北京大学数学科学学院
电 话: 010-62759090
摘要:In this talk, we will try to explain some of the ideas we have used in image analysis which include noise removal, image segmentation and image registration. We have used both TV-norm (total variation norm) regularization and level-set method for these problems. The talk will be divided into two parts.
In the first part, we try try to explain the essential ideas behind PDE (partial differential equation) methods used for noise removal. We will start with the linear removal methods used earlier and then come to the recent nonlinear filters including the well-known TV-filters. In the end, we will present a new 4th order PDE nonlinear filter. The advantages of the new filter will be explained and some of the other appraoches that are still under testing will also be explained.
In the second part, we will try to present the elementarty idea of the level-set methods. For some applications, we need to find the geometry of curves in 2D and surfaces in 3D to fit some criteria we impose. As a computational tool, level set methods are rather flexible and have proved to be able to capture rather complicated geometries in some design problems. In this part, I will summarize several joint works done with different colloborators recently about using level set methods and total variation regularization for inverse type of problems. The unifying theme is using total variation for regularization and using level sets for representing the geometry of discontinuities. Applications include elliptic inverse problems, positron emission tomography, electrical impedance tomography, digital image segmentation.
|
|