题目：Mathematical principles and image denoising

报告人：Quansheng LIU Universite de Bretagne -Sud (Univ. South Brittany), France

（E-mail: quansheng.liu@univ-ubs.fr）

Abstract：

In this talk, I will explain, with examples, how mathematical principles can lead to good improvements of classical filters in image denoising. Restoring a noisy image is a classical and important subject in image processing. An important progress was marked in 2005 by the introduction of the non-local means filters to remove the Gaussian noise. I will show how mathematical considerations, using probability theory, statistical estimation, and optimization techniques, can lead to a good comprehension and significant improvements of the non-local means filter and related algorithms. The improvements include in particular the optimization of the weights in the original filter for Gaussian noise, and the adaption of the filter to different kinds of noises (Poisson noise, impulse noise and mixture of noises).

时间：2019年12月20日（周五）下午16:00-17:00

地点：565net必赢客户端本部教二527教室

联系人：李海梁

主办单位：

565net必赢客户端

交叉科学研究院

北京成像理论与技术高精尖创新中心

欢迎全体师生积极参加！