学术报告-文有为

发布人:
责任人:
点击数:
更新时间:
2018-01-18 15:07:00

学术报告

题      目:Regularization Parameter Selection for Total Variation Based Image Restoration

报  告  人:文有为  教授  (邀请人:陈小山)

                        湖南师范大学数学与计算机科学学院


时      间:2018-01-18 10:45--11:30

地      点:学院401

报告人简介:

        主要从事科学计算、数字图像处理领域的研究。在SIAM J. Sci. Comput.、SIAM J. Imaging Sciences、SIAM Multiscale Model. Simul.、SIAM J. Matrix Anal. Appl.、Numer. Linear Algebra Appl.、IEEE Trans. Image Process.等期刊发表论文20余篇,目前论文他引次数超过300次。先后主持教育部留学回国基金1项、国家自然科学基金2项,参与香港研究资助局基金2项。


摘      要:

      The problem of image restoration is ill-conditioned. A total variation based regularization method should be used in the image restoration process. It is a very important task to select a suitable regularization parameter. By adjusting regularization parameter, a compromise is achieved to suppress the noise and preserve the nature of the original image. The appropriate compromise highly depends on the choice of the regularization
parameter. Usually, regularization parameter is determined manually by trial-and-error method, the generalized cross validation (GCV) method, the L-curve method, the discrepancy principle, etc. In this talk, we will show some results how to choose the regularization parameter for total variation based image restoration.