学术报告:高光谱图像去噪的秩逼近方法
主办单位:英国立博官网中文版
报告专家:李红,华中科技大学
报告时间:2018年9月21日(星期五)下午4:30
报告地点:数统学院201报告厅
专家简介:李红,教授,博士生导师,科技部国际科技合作计划评议专家,湖北省计算数学学会理事,美国IEEE会员。主要从事逼近与计算、机器学习与模式识别等方面的研究,在IEEE Trans等重要学术期刊上发表学术论文50余篇。主持国家自然科学基金、“十二五”航天支撑计划项目及国防预研基金等多个科研项目。2006年至2017年期间多次应邀访问香港浸会大学、澳门大学、美国加州大学尔湾分校(UCI)、澳大利亚悉尼大学等,十余次出席国际学术会议。2006年获宝钢教育基金“优秀教师”奖;2009年主持建设的“复变函数与积分变换”课程被评为国家精品课程,2013年评为国家精品资源共享课程;2013年获湖北省教学成果二等奖;2014年获湖北省名师称号。
报告摘要:Mixture noise removal is a fundamental problem in hyperspectral images (HSIs) processing that holds significant practical importance for subsequent applications. This problem can be recast as an approximation issue of a low-rank matrix. In this paper, a general non-convex smooth rank approximation (NCSRA) model is proposed to handle these mixture noises for hyperspectral images. The main idea is to use a general non-convex smooth function under some assumptions to directly approximate the rank function, which seeks a closer approximation than traditional methods.This non-convex optimization model can be easily solved by the convex analysis tooland remove the mixture noises of hyperspectral images quickly and effectively. Subsequently, we give a NCSRA iterative algorithm, and the corresponding convergence analysis is discussed and proved mathematically. Experimental results from simulated and real datasets illustrate that the non-convex smooth rank approximation method significantly outperforms the state-of-the-art methods on hyperspectral images denoising.