TUT-2: Accelerated Image Processing with FPGAs and GPUs

Date: Sunday Morning, October 12

Presented by

Shuvra S. Bhattacharyya, University of Maryland at College Park, Raj Shekhar, University of Maryland at College Park, William Plishker, University of Maryland at College Park, James Fung, NVIDIA Corporation, Joseph Stam, NVIDIA Corporation, Reinhard Koch, Christian-Albrechts-University of Kiel, Germany, and Christopher Zach, University of North Carolina at Chapel Hill

Abstract

This tutorial will focus on design and implementation of image processing systems based on field-programmable gate arrays (FPGAs) and graphics processors (GPUs). FPGAs and GPUs provide a unique mix of flexibility, programmability, and potential for streamlined acceleration. Both employ special purpose accelerators, memory structures, processor cores, and mechanisms for high-speed input/output to achieve significant performance benefits, but FPGAs and GPUs accomplish this in vastly different ways. FPGAs provide interconnections of programmable logic structures and interconnect that can be statically or dynamically configured to accelerate processing tasks in an application. GPUs are highly optimized arrays of processors capable of hundreds of gigaflops on commonly found graphics hardware and can be programmed in CUDA - an extension to standard C. This tutorial will review FPGA and GPU technology and methods for developing image processing applications, and will conclude with comprehensive case studies demonstrating the use of these high-performance architectures to accelerate computationally-intensive image processing algorithms. This tutorial will introduce NVIDIA CUDA, a C-language based programming environment for the GPU.

Speaker Biographies

Shuvra S. Bhattacharyya is a Professor in the Department of Electrical and Computer Engineering, University of Maryland at College Park. He holds a joint appointment at the University of Maryland Institute for Advanced Computer Studies (UMIACS). Dr. Bhattacharyya is coauthor or coeditor of four books and the author or coauthor of more than 100 refereed technical articles. His research interests include VLSI signal processing; biomedical circuits and systems; embedded software; and hardware/software co-design. He received the B.S. degree from the University of Wisconsin at Madison, and the Ph.D. degree from the University of California at Berkeley. Dr. Bhattacharyya has held industrial positions as a Researcher at the Hitachi America Semiconductor Research Laboratory (San Jose, California), and Compiler Developer at Kuck & Associates (Champaign, Illinois).

Raj Shekhar has academic and commercial experience in many types of medical imaging applications. He is presently an Assistant Professor of Radiology and Electrical & Computer Engineering and he has been a researcher and innovator in the field of medical imaging for over 10 years, during which time he has published over 30 refereed papers on the subject. He graduated from Indian Institute of Technology, Kanpur in 1989 with a B.Tech. in Electrical Engineering and went on to get his Masters in Bioengineering from Arizona State University. He recieved his Ph.D. from the Ohio State University in Biomedical Engineering. He also possesses 2-year medical imaging industry experience. Two of his inventions have led to successful commercial products (cardiac CT system marketed by Philips Medical Systems and intravascular ultrasound segmentation algorithms offered by Volcano Therapeutics). He holds three US patents.

William Plishker is a post-doctoral researcher at the University of Maryland focusing on application acceleration using different forms of parallism. He has published papers on application acceleration on multiple platforms including clusters, GPUs, FPGAs, and network processors. Most recently he has demonstrated orders of magnitude improvement to image registration using a combination of acceleration techniques on a heterogeneous platform. He graduated in 2006 from UC Berkeley with a Ph.D. in Electrical Engineering. Dr. Plishker's Ph.D research centered around the acceleration of network applications on network processors. He has worked at three different startups in Silicon Valley.

James Fung holds an M.A.Sc. and Ph.D. from the University of Toronto in Electrical & Computer Engineering. His work has been in the area of applying GPU Hardware for parallel general purpose computing, including implementing Computer Vision on the GPU. He is lead author of the OpenVIDIA project, which won the ACM Multimedia 2005 Open Source Software Award, and authored "Computer Vision on the GPU" in the popular GPU Gems 2 series of graphics programming books. He has authored or coauthored over a dozen peer reviewed papers in the area of parallel GPU Computer Vision and Mediated Reality. Merging technology with artistic endeavours, he created a system for the first Deconcert where 100 participants' brainwaves analyzed in real-time to drive music and sound synthesis. This work was featured on the Discovery Channel and Canadian national radio (Canadian Broadcasting Corporation, CBC). He currently works at NVIDIA examining computer vision and image processing on graphics hardware.

Joseph Stam is a Senior Applications Engineer for NVIDIA Corporation. He has a focus on computer vision applications of Graphics Processors, particularly in embedded applications. Prior to joining NVIDIA in 2007, Joe worked in the automotive industry for 12 years on research and development of imaging hardware and computer vision algorithms for vehicle based vision products. Joe received a B.S. degree in Engineering Physics & Computer Science from Hope College in Holland, Michigan and an M.S. degree in Electrical Engineering from Michigan State University. He is an inventor on 62 U.S. patents and several foreign patents, many of which relate to computer vision software and imaging hardware technologies.

Reinhard Koch holds a M.Sc. in Physical Science from the Fort Hays State University, Hays, Ks.,USA, and a Diploma Degree in Electrical Engineering from the University of Hannover. In 1996 he received his PhD (Dr.-Ing.) in Electrical Engineering in the field of stereoscopic 3D scene modeling from the Institute of Information Processing at the University of Hannover. From 1996 - 99 he lead the 3D modeling team at the KU Leuven, Belgium, in the vision group of prof. Luc van Gool. Since 1999 he is head of the Multimedia Information Processing group at the Department of Computer Science of the Christian-Albrechts-University of Kiel, Germany. The research interests of Reinhard Koch are 3D modeling from video and images, fast camera calibration and tracking algorithms, and the confluence of Computer Graphics and Computer Vision in the field of Mixed and Augmented Reality. Special emphasis is on real-time image processing and computer vision algorithms and sensor processing with the help of GPU hardware. He is involved in numerous large-scale research projects on 3D modeling, Augmented Reality, and 3D-Television, and author or coauthor of over 80 papers in the field. He teaches courses in Computer Graphics, Multimedia Technology and Compression, and Image-based 3D Reconstruction.

Jan-Michael Frahm is a research assistant professor in the 3D computer vision group at the University of North Carolina at Chapel Hill. He received his Diploma in computer science from Lübeck University, Germany. In 2005 he graduated from the Christian-Albrechts-University of Kiel, Germany, with a Ph.D. about sensor augmented camera calibration. The main focus of his research is reliable real-time 3D reconstruction from video or photo collections.