学术报告:Perceptual Coding: Hype or Hope?

报告1: Perceptual Coding: Hype or Hope?

【主讲人】 Professor C.-C. Jay Kuo

University of Southern California

【时 间】 2015年12月10日(周四)上午9:30-10:30

【地 点】 北京大学 计算机所大楼 106报告厅

Abstract:

There has been a significant progress in image/video coding in the last 50 years, and many visual coding standards have been established, including JPEG, MPEG-1, MPEG-2, H.264/AVC and H.265, in the last three decades. The visual coding research field has reached a mature stage, and the question “is there anything left for image/video coding?” arises in recent years. One emerging R&D topic is “perceptual coding”. That is, we may leverage the characteristics of the human visual system (HVS) to achieve a higher coding gain. For example, we may change the traditional quality/distortion measure (i.e., PSNR/MSE) to a new perceptual quality/distortion measure and take visual saliency and spatial-temporal masking effects into account. Recent developments in this area will be reviewed first. However, “is this sufficient to keep visual coding research vibrant and prosperous for another decade with such a modification?” The answer is probably not. Instead, I will present a new HVS-centric coding framework that is dramatically differently from the past. This framework is centered on two key concepts – the stair quality function (SQF) and the Just-Noticeable-Differences (JND). It will lead to numerous new R&D opportunities and revolutionize coding research with modern machine learning tools.

报告2: Deep Learning: Hype or Hope?

【主讲人】 Professor C.-C. Jay Kuo

University of Southern California

【时 间】 2015年12月10日(周四)上午10:45-11:45

【地 点】 北京大学 计算机所大楼 106报告厅

Abstract:

Deep learning has received a lot of attention in recent years due to its superior performance in several speech recognition and computer vision benchmarking datasets. A deep network can learn features (called deep features) automatically from training data. To understand deep learning, the first step is to understand these deep features. After a review of the short history of applying deep learning to vision applications, I will use two quantitative metrics to shed lights on trained deep features. They are the Gaussian confusion measure (GCM) and the cluster purity measure (CPM). The GCM is used to identify the discriminative ability of an individual feature while the CPM is used to analyze the group discriminative ability of a set of deep features. It is confirmed by experiments that these two metrics accurately reflect the discriminative ability of trained deep features. Further studies with the metrics as tools reveal important insights into the deep network, such as its good detection performance of some object classes that were considered difficult in the past. Finally, I will explain my view to the deep learning methodology - its pros, cons and future perspectives.

Biography:

Dr. C.-C. Jay Kuo received his Ph.D. degree from the Massachusetts Institute of Technology in 1987. He is now with the University of Southern California (USC) as Director of the Media Communications Laboratory and Dean’s Professor in Electrical Engineering-Systems. His research interests are in the areas of digital media processing, compression, communication and networking technologies.

Dr. Kuo was the Editor-in-Chief for the IEEE Trans. on Information Forensics and Security in 2012-2014. He was the Editor-in-Chief for the Journal of Visual Communication and Image Representation in 1997-2011, and served as Editor for 10 other international journals. Dr. Kuo received the National Science Foundation Young Investigator Award (NYI) and Presidential Faculty Fellow (PFF) Award in 1992 and 1993, respectively. He was an IEEE Signal Processing Society Distinguished Lecturer in 2006, and the recipient of the Electronic Imaging Scientist of the Year Award in 2010 and the holder of the 2010-2011 Fulbright-Nokia Distinguished Chair in Information and Communications Technologies. Dr. Kuo is a Fellow of AAAS, IEEE and SPIE. Dr. Kuo has guided 130 students to their Ph.D. degrees and supervised 23 postdoctoral research fellows. He is a co-author of about 230 journal papers, 870 conference papers and 13 books.

CLOSE

上一篇 下一篇