好色tv

报告主题：

1. Graphical Model-based View Synthesis Distortion Estimation

2. Approximate Message Passing-based Compressed Sensing Reconstruction with Generalized Elastic Net Prior

报告主讲：梁杰 教授 加拿大西蒙弗雷泽大学

报告时间：2014-07-08（周二） 09：30-11：30

报告地点：九教北609会议室

主办单位：研究生工作部

承办单位：计算机与信息技术好色tv

【报告人简介】

Jie Liang received the B.E. and M.E. degrees from Xi'an Jiaotong University, China, the M.E. degree from National University of Singapore (NUS), and the PhD degree from the Johns Hopkins University, Baltimore, Maryland, USA, in 1992, 1995, 1998, and 2003, respectively. Since May 2004, he has been with the School of Engineering Science, Simon Fraser University, Canada, where he is currently an Associate Professor. In 2012, he visited University of Erlangen-Nuremberg, Germany, as an Alexander von Humboldt Research Fellow. From 2003 to 2004, he worked at the Video Codec Group of Microsoft Digital Media Division. From 1997 to 1999, he was with Hewlett-Packard Singapore and the Center for Wireless Communications (now part of the Institute for Infocomm Research), NUS.

【报告简介】

1. Graphical Model-based View Synthesis Distortion Estimation

In Depth-image-based rendering (DIBR), a virtual view is synthesized from the textures and depths of the reference views. In this talk, we consider the two DIBR algorithms used in the MPEG View Synthesis Reference Software (MPEG-VSRS), and study the impact of the transmission error in the reference views on the quality of the synthesized view. A graphical model is developed to capture the relationship between the reference views and the synthesized view, including the warping competition operation in the DIBR algorithms. The probability mass function and the expected distortion of each synthesized pixel are then calculated in the presence of packet loss. Experimental results verify the accuracy of the proposed method.

2. Approximate Message Passing-based Compressed Sensing Reconstruction with Generalized Elastic Net Prior

In this talk, we consider the compressed sensing reconstruction problem with an additional estimation of the signal available during the reconstruction, which can be viewed as a generalized elastic net prior. The powerful Approximate Message Passing (AMP) algorithm is first generalized to incorporate the elastic net prior. Next, based on the corresponding state evolution formula, the asymptotic prediction performance and noise sensitivity of the scheme are derived. Finally, a practical parameterless AMP algorithm is developed. Simulation results are then presented to verify the efficiency of the proposed method.