Stanford News


CONTACT: Stanford University News Service (650) 723-2558

Image science gaining new emphasis

The old world of images: analog, black and white, office-centered, stand-alone, passive.

The new world of images: digital, color, home-centered, communicative, interactive.

This sweeping change in the way that the world around us is represented is the impetus behind Image Systems Engineering, a multidisciplinary program established last fall by professors from psychology, electrical engineering and computer science. Their aim is to train students for the growing number of industry jobs in the new field of imaging science.

"The science and engineering of images has matured to the point where it makes sense to combine them into a single curriculum," says Brian Wandell, the professor of psychology who was the prime mover behind the program's formation.

The combination of the three fields makes perfect sense, says psychologist David Heeger, who also has courtesy appointments in both computer science and electrical engineering. Imaging science is a peripheral subject in all three traditional disciplines, Heeger says. In psychology, it is the study of human vision. In computer science, it is research into computer vision. And in electrical engineering it is image processing. "Together they form quite a coherent discipline, but very few places approach these fields in such an integrated fashion," he says. Tangible evidence of industry interest in the area is the $1 million equipment grant that the Hewlett-Packard Corporation gave to the program last fall to equip a new image processing laboratory in the basement of Margaret Jacks Hall. The company provided 15 graphics workstations with a full range of peripherals, including high-speed color scanners, printers and digital cameras. The start-up effort has also received financial support from Xerox Corp.

Hewlett-Packard had circulated a request for proposals on programs in image systems engineering. Out of 23 applications, the company funded five. In addition to Stanford, the University of California-San Diego, the Georgia Institute of Technology, the University of Iowa and Purdue were selected. "We wanted to fund the real thought leaders in the area, which is critical to the future of our business," said Nancy Levitt, H-P University Grants Program manager.

Rochester, N.Y., has long been the center of image science in the country, due mainly to the influence of Kodak and Xerox, says Joseph Goodman, professor of electrical engineering and senior associate dean in the School of Engineering, who has supported the new program. Research in Rochester has focused primarily on print and photography. As H-P's interest suggests, Silicon Valley companies are becoming increasingly involved and interested in digital image processing. So it is an appropriate area for Stanford to develop, he said.

"We've got the number-one ranked EE department in the country, the number-one ranked computer science department, and the number-one ranked psychology department. What a combination," Goodman says.

In the summer of 1985, after Wandell came to him with the idea of forming the program, Goodman organized a lunch for faculty members who are working in this general area. Among those who attended were Ronald Bracewell, Tom Cover, Robert Gray, Teresa Meng and Dwight Nishimura from electrical engineering; Patrick Hanrahan, Mark Levoy and Carlo Tomasi from computer science; Heeger from psychology; and Richard Olshen from Biostatistics. Since that initial meeting, Parvati Dev from the Medical School, Abbas El-Gamal from electrical engineering and James Gibbons, special counsel in the President's Office, have also become involved.

"I was really surprised at the large number of people we have who do something in imaging. I knew them all, but had never seen them gathered in a single room before," said Wandell. The group agreed to combine their efforts and adapt the various imaging-related classes that they teach to take advantage of the new laboratory and to create an integrated curriculum.

The curriculum now includes 14 courses that fall into three paths: medical imaging; image processing; and Fourier and statistical optics. In the program's first year about 90 students have taken courses that have utilized the lab, Wandell said.

One of the unusual aspects of the new program is the heavy involvement of the psychologists. This makes perfect sense, said Wandell, when you consider the fact that all color reproduction is an illusion. "If you measure the light reflected from a painting and the light emitted from an image of the painting displayed on the CRT, the two look totally different. But to the human eye they look nearly the same." So to make high quality images it is imperative to understand how human vision works, and that is a subject psychologists have been studying for more than 125 years.

"It's a good thing to have engineering students [who are] interested in imaging learn about human vision early in their careers," said Xuemei Zhang, a fifth-year graduate student in psychology who served as an unofficial teaching assistant in Wandell's course. After a number of years of academic study, Zhang was attracted to image systems "because it allows me to do something practical with what I have learned." She is one of the new program's first success stories, having accepted a job at H-P to develop vision models that can predict the quality of printed and displayed images.

The new program and lab have really improved the image systems courses, Zhang said. "When I first took Brian's course [Psych 221/EE 362] it was pretty much a lecture and exercise course. It is much better now as a laboratory, project-oriented course. Students can apply what they learn to specific problems."

The well-equipped lab has made a new level of learning possible, Zhang said. The equipment is good for getting students involved. They can, for example, check out the digital camera and learn about the problems of the technology firsthand. Also, the powerful workstations provide quick feedback that allows the students to get more done.

Xin Tong is a doctoral student in electrical engineering whose primary interest is in image compression technology. She also maintains the program's website. She was attracted to image processing because "you can see how it is useful because of the explosion of the Internet and multimedia. And, of course, it's cool. It's much more pleasant to look at images, rather than circuit diagrams: I look at the results of what I do and I am pleased."

Before the program was set up, the information about image systems was "sort of spread out around campus," she said. "There were professors here and there, and people didn't know about them. Now you can look at the entire field and pick your specialization."

Goodman would like to see Image Systems Engineering evolve into a program comparable to the Center for Telecommunications, which involves faculty from several departments, provides extra visibility to the area, and draws on industry to support research and to help identify problems to work on. "With a modest investment of billets for strategic appointments to fill in a few gaps in faculty expertise, we could become a real powerhouse in this area," he said.

For more information on the World Wide Web contact:
The Image Systems Processing Lab home page


By David F. Salisbury