Computer Vision and Multimedia Communications
3D Computer Vision & Graphics, Multimedia, Networks.
The focus of my present academic research is in Quality of Service (QoS) based delivery of Multimedia for Electronic Commerce and TeleLearning. For both these applications it is necessary to adaptively use the bandwidth available and provide the best possible quality according to user priority specifications. I have developed a statistical approach for "Optimal Bandwidth Monitoring" for single server connections as well as for distributed multimedia retrieval. The approach provides an estimation based on statistical confidence level and has shown to produce expected results based on real network tests.
In the context of online 2D & 3D visualization, I am developing JAVA Applet based tools for interactive viewing of Super High Resolution & 3D images; this work is being supported by TelePhotogenics Inc. and the National Research Council of Canada. More importantly, I raised substantial industrial funding to build a Super High Resolution digital camera and have designed patented and patent pending technologies for new SHR Stereo and 3D scanners.
I have also been active in the Synthetic Natural Hybrid Coding (SNHC) component of MPEG-4 coding. SNHC is part of Working Group 11 (WG11) in MPEG-4. Specifically, I have been working on detection and tracking of facial features, and displaying such information in a videoconferencing scenario through model-based coding and transmission. I am also investigating displaying head-and-shoulder models in stereo. This research will be further supported by the installation of a major equipment facility, the "CAVE", which uses rear projection of stereo images on to 3 walls.
In the past, I initiated the applications of variable resolution (or foveation) emulating the animate visual systems for image compression and communication. Results have shown how intelligent pre-processing of images can not only lead to improved videoconferencing systems but can also enhance the quality of standard compression algorithms, such JPEG, MPEG, and Wavelets. The results are especially useful for communication at low bit-rates. I have also demonstrated the advantage of "intelligent" cell loss for image/video data transmitted over congested ATM networks. The work will have an influence on the emerging area of Multimedia ATM design. My publications clearly shows that the traditional approach of treating all types of information as just a stream of bits in ATM switches is inadequate for congested networks. My work on foveated image compression and stereo visualization has been referenced and extended by several groups of researchers at New York University, Simon Fraser University and University of Texas at Austin.
I have been the lead applicant in several projects including an Imaging Systems Curriculum initiative funded by Hewlett-Packard with $365,000 in equipment & a recent ASRA/TelePhotogenics/IBM/Keeweetinok Lakes RHA 3D Medical Imaging initiative that has received over $2 M in cash and in-kind funding.
While traditional image and video remain at the core of multimedia content, 3D video is perceived as the next generation in video technology. 3D video incorporates the depth perspective which enables viewers to feel immersed in a more realistic environment. This course provides students with the latest 2D and 3D video developments and in particular relating to stereoscopic and multi- view with or without special eye-wear. Many of the techniques proposed on 3D video inherit much of the strengths from 2D video methods and computer vision techniques. The 3D component is also included in the latest HEVC standard. This course will focus on literature review and survey of these techniques. Group studies, discussions and presentations constitute the main thrust of the course. Sections may be offered at an increased rate of fee assessment; refer to the Tuition and Fees page in the University Regulations sections of the Calendar.Winter Term 2023
Artificial Intelligence (AI) in multimedia covers a wide range of topics. In general, it means simulating human intelligence using computer algorithms. This course introduces a high level understanding of machine learning/deep learning, which is a branch of AI. The instructor may decide to include reinforcement learning and other aspects of neural networks, as well as natural language processing depending on the course schedule. Sections offered at an increased rate of fee assessment; refer to the Tuition and Fees page in the University Regulations sections of the Calendar.Fall Term 2022