1066vip威尼斯党委宣传部、新闻中心主办

数学学科2024系列学术报告之四十一

来源:理学院 发布日期:2024-12-17

  题目:Low-rank regularization methods for hyperspectral and multispectral image fusion

  报告人: 张俊

  时间:2024年12月19日(周四),上午15:00-16:00

  地点:理学院1-301会议室

  报告摘要: Recent research has highlighted the effectiveness of nuclear norm in addressing the fusion of Hyperspectral Image (HSI) and Multispectral Image (MSI) in the same scene. However, the standard nuclear norm method fails to differentiate between different singular values during processing, leading to certain limitations and shortcomings in practical applications. To address this issue, this report investigates HSI-MSI fusion methods from two perspectives: matrix decomposition and tensor decomposition: (1) Innovatively introducing the concept of weighted nuclear norm from image denoising to ensure the preservation of critical data components during image fusion. Specifically, a unified framework integrating weighted nuclear norm, sparse prior, and total variation regularization is proposed; (2) To deeply explore the low-rank characteristics of HSI, this report introduces a newly developed HSI-MSI fusion method within the framework of Tensor Ring (TR) decomposition by integrating the TR factor-based logarithmic tensor nuclear norm with weighted TV.

  报告人简介:

  张俊,南昌工程学院理学院副教授,硕士生导师,江西省优秀青年基金获得者。2013年6月获湖南大学理学博士学位,并在该校电气与信息工程学院进行了博士后研究工作。2017.10-2018.10美国德克萨斯大学访问学者,并于2024年短期访问香港城市大学。现为“应用统计”硕士专业学位点数据科学方向的负责人,江西省电子学会理事。主要研究方向:高光谱遥感图像处理,数值最优化,图像复原与分割。主持在研国自科地区科学基金项目、江西省自然科学基金优秀青年基金项目和面上项目各1项;主持完成国家级、省级科研项目4项。在IEEE TGRS、IEEE JSTARS、SP、AMC、AMM等著名学术期刊上发表学术论文30余篇,获得授权发明专利1项。