On the Theoretical Convergence of Plug-and-Play Iterations

  报告人:Kunal Chaudhury

  报告摘要:The topic of this talk is a recent paradigm for image reconstruction called the Plug-and-Play (PnP) method which has attracted lot of attention for computational imaging problems. This involves repeated inversion of the imaging (forward) model followed by ad-hoc regularization using a powerful denoiser. Surprisingly, though PnP is not derived from a statistical or variational perspective, it yields state-of-the-result in several imaging problems. In fact, the theoretical aspects of PnP, such as convergence and optimality, is not fully understood even for linear denoisers.  In this talk, we will discuss some recent progress in this direction and the open problems at this point.

  报告时间:2021年1月14日17时

  报告地点:腾讯会议 (ID:143 142 066)

  主办单位:数学科学学院

  报告人简介:

  Dr. Kunal Chaudhury (PhD, EPFL Switzerland) is an Associate Professor in the Department of Electrical Engineering, Indian Institute of Science, where he heads the Lab for Imaging Sciences and Algorithms. His research interest is in image processing, computational imaging and computer vision, with a focus on numerical optimization and fast approximations. He has worked in areas ranging from wavelet algorithms to nonlinear filtering, and from regularization methods in imaging to semidefinite programming for sensor network localization. He is a Senior Member of IEEE, a member of SIAM and IEEE Signal Processing Society, and a Senior Editor of the SPIE Journal of Electronic Imaging.

数学科学学院 聊城大学西校区11号教学楼
办公电话:8238817 邮箱:math@lcu.edu.cn
Copyright © 2022 math.lcu.edu.cn , All Rights Reserved.