Abstract:Distances between images or videos are crucial for lots of applications, such as event recognition and near duplicate image identification. Recently, multi-level matching methods were proposed for efficient distance computation and demonstrated promising results in different tasks. In this talk, I will introduce two recent works from our group on aligned pyramid matching in spatial and temporal domains respectively. In both methods, we divide the images (or videos) into subimages (or subvideos) at multiple levels. At each level, we propose to use two matching stages to efficiently compute the distances. Finally, we fuse the information from multiple levels. The exhaustive experiments demonstrated promising performances of both methods in two applications.

Bio: Dong Xu  is currently an assistant professor at Nanyang Technological University at Singapore. He received the B.Eng. and PhD degrees from the Electronic Engineering and Information Science Department, University of Science and Technology of China, in 2001 and 2005, respectively. During his PhD study, he worked at Microsoft Research Asia and The Chinese University of Hong Kong for more than two years. He also worked at Columbia University for one year as a postdoctoral research scientist. His research interests include computer vision, pattern recognition, statistical learning and multimedia content analysis.

He has published more than 20 papers in top venues including T-PAMI, T-IP, T-CSVT, T-SMC-B and CVPR. He is an associate editor of Neurocomputing (Elsevier). He is the guest editors of three special issues on video and event analysis in IEEE Transactions on Circuits Systems for Video Technology (T-CSVT), Computer Vision and Image Understanding (CVIU) and Pattern Recognition Letters (PRL), and a coauthor of a forthcoming book entitled "Semantic Mining Technologies for Multimedia Databases". He was awarded a Microsoft Fellowship in 2004.