Date: 2010-11-22 00:00

Speaker: Pradeep Sen

Abstract:
Compressed sensing is a fast-growing field in the applied math community that shows how to reconstruct signals from a small set of random samples if the signals are sparse in a transform domain. Although the theory has engaged researchers in many fields because of its exciting potential applications, there is unfortunately a gap between the elegant theory and practice because of its less-than-stellar performance on real-world data sets and the lack of robustness of the real algorithms. At the UNM Advanced Graphics Lab, we have been working to address these practical shortcomings and are rapidly becoming one of the leading research centers for compressed sensing applications in computer graphics and imaging. In this talk, I will describe several major innovations that we have developed that allow us to apply compressed sensing to real problems. First, I will propose a novel algorithm for efficient light transport acquisition based on our earlier work on dual photography. This algorithm takes advantage of the compressibility of light transport to accelerate the acquisition of a 4-D image-based data set. We demonstrate that this results in a very efficient method for capturing the light transport. Next, I will develop the foundations of two new algorithms that enable us to apply compressed sensing to the problem of rendering images. In this work, we present the principle of “compressive rendering” wherein the sparsity of the final image is used to accelerate its calculation. We demonstrate results with our framework integrated into a Monte Carlo rendering system.

Bio:
Dr. Pradeep Sen is an Assistant Professor in the Department of Electrical and Computer Engineering at the University of New Mexico with joint appointment in Computer Science. He received his B.S. in Computer and Electrical Engineering from Purdue University in 1996 and his M.S. in Electrical Engineering from Stanford University in 1998 in the area of electron-beam lithography. After two years of working at a profitable startup company which he co-founded, he joined the Stanford Graphics Lab where he received his Ph.D. in Electrical Engineering in June 2006, advised by Dr. Pat Hanrahan. After arriving at UNM in the Fall of 2006, he founded the UNM Advanced Graphics Lab, which is dedicated to research into the science and algorithms of computer graphics, visualization, computer vision, and image processing. His research interests include real-time graphics and graphics hardware, global illumination algorithms, computational photography, and applications of signal processing theory to computer graphics. His research has been featured on Slashdot, New Scientist, CGWorld, CPU, Technology Research News Magazine, and even on the popular TV show Numb3rs. Dr. Sen was recently awarded an NSF CAREER award in 2009 to study the application of compressed sensing theory and sparse reconstruction algorithms to computer graphics and imaging.