Qiong Liu
Assistant Researcher   
Broadband Network & Digital Media
Lab, Department of
Automation, Tsinghua University, Beijing 100084, China 
Telephone:0086-10-6278-8613
ext 816 (o)
E-Mail:  
qiongliu@tsinghua.edu.cn
Education Background 
・        
Ph.D  Sep. 2003 ~ Dec.2008,
National Engineering Research Center for Multimedia Software, Wuhan University,
Wuhan, China. 
・        
B.S.  Sep. 1999 ~
July 2003, Dept. of Computer Science and Technology, Central China Normal
University, Wuhan, China
Professional Experience
Apr.2010,  Assistant
Researcher, Broadband Network &  Digital Media Lab, Tsinghua
Univeristy, Beijing, China
Aug. 2009, Lecturer,
Dept. of Electronic and Information Engineering, Huazhong University of
Science & Technology, Wuhan, China
Research Interests
Multi-view Video Coding
Depth Estimation
for Multi-view Video
Low Complexity Video Coding
Region-of-interest Video
Compression
Scalable Video Coding 
Video Surveillance
Research
Projects
Joint Efficient Multi-view
Video Coding and Depth Generation for Interactive 3D Video
There is an increasing demand for
three-dimensional (3D) video since it provides more real perception, interactivity
and novel visual enjoyment. The emerging interactive 3D applications can not
only bring 3D depth impression of a scene, but also allow users an
interactive selection of the viewpoint. However, multiview video captured
simultaneously from multiple cameras at slightly different view or angles
requires much more storage space, transmission bandwidth and computational
resources. Instead of that, techniques of virtual view synthesis can produce
virtual views based on depth information and real views in neighbors.
Therefore, the number of views have to be compressed and transmitted is
reduced obviously. One of the key points of 3D system is depth generation. It
requires high quality and low-complexity. To make a solution for these
problems, a joint efficient multi-view video coding and depth generation
scheme is proposed. Aiming to improve the depth quality, motion analysis and
object segmentation are implanted based on color video information and the
results are very helpful for depth generation. For the issue of complexity,
depth and disparity estimation lead to high computation complexity while
strong correlation is retained between depth and motion information. Thus, it
is worthy to design a joint depth and disparity estimation approach to reduce
the complexity in encoder. Besides these, the stereo perception is also
needed to investigated, which could highly improve the perceptual quality of
3D Video.
Low Complexity Coding Algorithm
Based on Interactive 3D Video
To achieve high compression efficiency of
encoding multiple view data, time-consuming technologies, including variable
block size Motion Estimation (ME)/Disparity Estimation (DE) and
Multi-Reference Frame (MRF) selection, are adopted in MVC. Consequently,
tremendous complexities are induced and it leads to difficulties for
interac