Zhu Li's Homepage at Hong Kong Polytechnic Univ


About me:

I am now an Asst. Prof. with the Dept of Computing, Hong Kong Polytechnic University. Before that I was with the Multimedia Research Lab, Motorola Labs , USA, from 2000~2008, where I was a Principal Staff Research Engineer. I received my PhD in Electrical & Computer Engineering from Northwestern University, Evanston, USA, in 2004.

I am an IEEE Senior Member, elected Vice Chair for IEEE Multimedia Communication Tech Committee (MMTC), and my recent awards include a Best Poster Paper Award from IEEE International Conf on Multimedia & Expo (ICME), 2006, Toronto, and Best Paper Award from IEEE International Conf on Image Processing (ICIP), 2007.

For more info: my Bio, CV,, and old webpage .
Contact: Office: PQ-714, Office Phone: (852) 2766 7316, Email:

Teaching:
Fall 2009, COMP100 Intro to Info Tech
Spring 2009, COMP435 Biometrics
Fall 2008, COMP212 Computer Architecture

Have Fun:


Research Summary:

My research interests are in video coding and communication, optimization and distributed computing techniques in multimedia streaming and networking, video signal analysis and machine learning with applications in biometrics, video event recognition, as well as large respository video search and mining.

My research projects are currently supported by grants from Microsoft Research Asia, Hong Kong RGC/GRF, and PolyU, which are graciously acknowledged here.

Call for Chapters: Intelligent Multimedia Communication: Techniques and Applications , Ed. Changwen Chen, Zhu Li, and Shiguo Lian, Springer-Verlag, 2009.
Call for Chapters: Multimedia Analysis, Processing and Communication, Ed. Z. Li, J. Kacprzyk, D. Tao, W. Lin, E. Izquierdo, and H. Wang, Springer-Verlag, 2009.

Research Group

Post-Doc/PhD/Research Assistant Recruiting: support is available for motivated students in areas of video analysis, video search and mining, video communication and networking. Please see my related research projects for more detail. See the current opening for more detail.

Some highlights of recent projects and selected publications (book chapters/journal papers and submissions in green):

Multimedia Communication:

Adances in video signal processing and coding give us a rich set of video coding and adaptation tools with associated quality metrics, how to utilize these tools and metrics, and integrate with underlying network engineering elements like channel and network coding, routing and resource allocation and optimization solutions, developing a distributed, scalable and adaptive content delivery network solution, are my interests.

[1] Internet Video Delivery: utility gradient driven scheduling, P2P, content-aware, source-channel coding, elasticity and R-D optimization

[2] Multi-Access Multimedia Networking: pricing model on resource allocation, distributed coordination for outer loop control, while source coding/adaptation R-D optimization in the inner loop:

[3] Wireless Video: video source adaptation, cross-layer optimization, source-channel coding, resource allocation and collaboration, local relays, wireless P2P, energy efficiency.

[4] Video Summarization & VLBR Video Streaming: Frame drop distortion metrics, Viterbi algorithm based frame drop optimization to minimize frame drop distortions, with applications in VLBR video streaming (e.g. QCIF "foreman" sequence streaming at 18kbps, demo available on request).

Multimedia Computing:

Currently I am interested in spatio-temporal appearance modeling, piece-wise linear approximation of non-linear appearance manifolds, with query driven and/or global structures, and their applications in multimedia computing problems.

[1] Visual Pattern Recognition and Biometrics: appearance manifold modeling, local embeddings, diffusions, model localization with piece-wise linear model approximation of a global non-linear manifold, with application in face recognition, head-pose estimation and video search metrics. [2] Image/Video Search and Mining : repeated clip search and mining with the LUminance Field Trajectory (LUFT) modeling, scalability in searching large video repositories, SIFT based image similarity search. Our LUFT based repated video clip searching achieves very high performance in speed (0.012sec to search an 5-hour collection) and precision-recall (100% vs 96%), see a report,
updated: 10/10/2009, by Z. Li.