|NSF Information Technology Research
Steven M. Nowick, foreground, and Kenneth L. Shepard.
A quartet of Columbia Engineering faculty will be embarking on long-term, risk-taking research, having received $6.5 million from the National Science Foundation as part of its $90 million initiative in Information Technology Research (ITR). In announcing these first ITR awards, President Bill Clinton noted that high technology accounts for one-third of the U.S. economic growth in recent years and generates jobs that pay 85 percent more than the average private sector. Columbia was one of five universities receiving the most large-scale ITR awards.
The largest of Columbia's grants for cutting-edge research, $3.5 million, was awarded to Shree Nayar, professor and acting chair of computer science, who will be concentrating on visual information processing. Steven M. Nowick, associate professor of computer science, will be principal investigator on a $1.6 million, five-year grant to produce a computer-aided design (CAD) framework for large-scale asynchronous digital systems. He also will be working with Kenneth L. Shepard, associate professor of electrical engineering, who is principal investigator on a three-year, $970,000 grant to create very energy-efficient, high-performance programmable digital signal processors for third-generation wireless systems.
News of the funding of Dr. Nayar's proposal arrived shortly after the announcement of his most recent invention, spacially varying pixel exposures. This technique enhances the range of brightness of any imaging system - motion picture film, video, photography, magnetic resonance, X-ray or infra-red. The second step of the technology uses algorithms to reconstruct high-quality images from the image captured using the spatially varying pixel exposure technology. The result is images comparable to those produced by a much more sophisticated digital camera.
Dr. Nayar's NSF project is broader in scope than any of his previous research. "It is an ambitious goal to attempt a higher degree of photo-realism," he said. "We expect to capture a minimum amount of visual information, either photographic or video, and understand, with mathematical models of illumination and geometry, how to predict new appearances of the scene. The scene can be in a new illumination condition with differing viewpoints and we will be able to remove or replace 3D objects," he continued.
As principal investigator, Dr. Nayar will head a team of scientists from five other universities to develop technical tools to manipulate visual data. The project encompasses computational vision and computer graphics, with applications in industry, entertainment, architectural design and many other fields. He explained:
"We expect to develop tools that will allow us to interact with and create variations of scene appearances. For example, if we have a photograph of an automobile, we can get other views or simulate a new model and interact with it by varying illumination, or changing color, or changing the view, all through software. For animated movies, we could produce characters that look more realistic, with skin and hair and a richness of visual detail not possible now. In construction, we will be able to take an architectural design and predict how it will look after renovation or how it might age or corrode. With the Internet, our goal is to have 3D environments through which one can navigate, similar to being in an art gallery where you move yourself through a space to determine where you want to go."
The focus of the new NSF grants emphasizes human-computer interfaces, such as Dr. Nayar's, as well as revolutionary computing and software development. Drs. Nowick and Shepard's research can be called revolutionary. In discussing their joint research efforts, the two colleagues described the underlying theory that drives both their joint and individual research, each elaborating on the other's statements.
"It is a paradigm shift from synchronous to asynchronous," said Dr. Nowick, "For 50 years, the synchronous clock has been used for computing, but it has hit its limits and so, in looking at alternatives, asynchronous (clockless) design is promising in solving the problem."
Dr. Shepard said the problem is obtaining high performance in programmable digital platforms at radically lower power. "Synchronous means a chip with a clock, but this will be a chip with no clock," he said. "Asynchronous techniques will enable the aggressive use of energy-saving techniques, such as very aggressive, software-controlled scaling of the supply voltage, while preserving high performance where it's needed. The Holy Grail is a software radio in which nearly all of the signal-processing is performed digitally on a programmable platform. It was a nice synergy between my research in advanced design techniques for deep submicron CMOS and Steve's work in asynchronous design."
Working with a colleague at USC, Dr. Nowick will develop an automated computer aided design framework that will overcome limitations of current asynchronous design tools. An important goal of this project is "design-space exploration," creating new optimization algorithms and software tools that allow the designer to try out alternatives and find the best design for a particular application. They also will explore the possibility of a synthesis of synchronous and asynchronous systems.
"Ken and I will design a chip," said Dr. Nowick, "and, in the project with USC, we will produce software tools, but we will also have a chip." Prof. Nowick is one of four investigators nationally to be awarded two large-scale grants.Steven K. Feiner, professor of computer science, will use his three-year grant of $450,000 to investigate the design of user interfaces that combine together different kinds of displays and interaction devices. For example, multiple mobile users may view a shared wall-mounted data visualization. At the same time, each user might see complementary private material, customized to her own information needs, and overlaid on the common display through the use of head-tracked hand-held or head-worn see-through displays that create an "augmented reality." "I'll be exploring what I call 'environment management'," said Dr. Feiner. "This is the problem of creating effective user interfaces that allow people to interact with a large and changing mix of personal and shared displays, without becoming overwhelmed by low-level details."
NSF received 920 pre-proposals for large projects, of which 62 were funded. "The competition for these ITR grants was significant," said Dean Galil, "and our faculty did very well."
Return to front page
Return to SEAS Alumni