学术报告:Graph Signal Processing for Image Compression and Restoration

报告题目:Graph Signal Processing for Image Compression and Restoration

报告人:胡玮 博士

时 间: 2016年11月29日(周二)下午1:30-3:00

地 点: 北京大学计算机科学技术研究所1层106会议室

Abstract: Graph Signal Processing (GSP) is a fast-growing field for the processing of signals defined on graphs—graph signals. Graph signals include signals that naturally reside on graphs, such as posts on social networks, and signals whose structure can be embedded into graphs, such as images. A wealth of tools have been developed in GSP to address multimedia applications. In this talk, I will introduce our works on GSP for the compression and restoration of images. First, I will explain the benefits and challenges of GSP. Then, I will focus on two works on image compression based on Graph Fourier Transform defined in GSP and its generalization. The first work is multi-resolution Graph Fourier Transform for the compression of a class of images called piecewise smooth images, in which we prove the optimality of Graph Fourier Transform and design optimal transform coding accordingly. The second work extends Graph Fourier Transform to Generalized Graph Fourier Transform by considering boundary conditions, where we show the optimality of this transform and apply it to the compression of image prediction residual. Both works achieve state-of-the-art compression performance.

Bio: Wei Hu received the B.S. degree in electrical engineering from the University of Science and Technology of China in Jun. 2010, and the Ph.D. degree in electronic and computer engineering from The Hong Kong University of Science and Technology in Aug. 2015. She has been a researcher in Imaging Science Lab in Technicolor, France since Sept. 2015.

Her research interests include image/video processing and compression, graph signal processing, and computational photography. She has served as a journal reviewer for IEEE Transactions on Image Processing, IEEE Transactions on Circuits and Systems for Video Technology and IEEE Signal Processing Letters, and a conference reviewer for ICIP, ICME, MMSP, etc.

[关闭]
版权所有 © 2007 北京大学计算机科学技术研究所