Light Field Duality, Uncertainty and Free Viewpoint Video

Speaker:          Dr. George Chen   
                  STMicroelectronics San Diego Laboratory    

Title:            Light Field Duality, Uncertainty and Free Viewpoint Video



Date:             Monday, 29 March 2004 

Time:             4:00pm - 5:00pm 

Venure:           Lecture Theatre F   
                  (near lift nos. 25/26)   


ABSTRACT: 

In recent years, Free Viewpoint Video (FVV) has attracted the attention 
of the international MPEG committee. Concrete standardization steps 
towards supporting FVV have been taken. FVV aims at providing viewers 
both stereo parallax and motion parallax. Depending on the rendering 
method, three coding technologies have been proposed based on depth, 
geometry model or light field. I started studying light field duality 
in late 2002, with the purpose of replacing the then state-of-the-art 
signal processing approach which applies only to uniformly sampled light 
fields. Based on the duality formulism, the equivalence between the
geometry-based and light field-based rendering methods can be rigorously 
proved. This paves the way for a unified FVV codec. In this presentation, 
I will introduce the duality concept and show how it leads to an elegant 
rendering algorithm that can be accelerated by off-the-shelf graphics 
hardware. I will also touch upon my recent proposal to MPEG about a 
unified multi-view coding approach. In practice, light fields cannot be 
exactly sampled. My colleagues and I have developed methods and 
representations to estimate uncertainties associated with light field 
samples. I will show how the uncertainty information can be used to 
generate probably correct novel views.

BIOGRAPHY: 

Dr. George Chen is currently a research manager at STMicroelectronics 
San Diego Laboratory. He obtained his Ph.D. degree in Computer Science 
from the University of Southern California in 2000. Prior to joining 
STMicroelectronics, he was a consultant to Geometrix Inc helping the 
company create a 3-D face recognition system based on his doctoral work. 
Dr. Chen is interested in many aspects along the chain of visual 
information sensing, processing, transmission and display. He is also 
interested in parallel microprocessor architectures to support real-time
visual signal processing.