PAL-2 Computer Promises Dramatic Improvements In Image Processing

GAINESVILLE, Fla. --- A pizza-box size desktop computer as fast as 9,000 Pentium processors soon will dramatically speed and improve everything from airport luggage checks to mammograms, using a mathematical computer shorthand developed at the University of Florida.

Known as PAL-2, for Parallel Algebraic Logic, the computer also will bring about equally dramatic drops in the cost of image processing, said Gerhard Ritter, professor of computer engineering and mathematics and director of UF's Computer Vision and Visualization Research Center. Ritter, along with fellow UF researcher Joseph Wilson, developed and pioneered the mathematical theory known as image algebra that makes the technology possible.

"When we began developing image algebra, we wanted to provide the military and the commercial market with an image processing capability that would significantly reduce the cost of the next generation of high-speed image and signal processors," Ritter said. "By the year 2000, thanks to image algebra and the PAL-2 chip, the cost of powerful imaging computers will come down from today's half a million dollars to about $4,000."

The chip design for an evolved PAL-2, a 4-by-5-inch multi-chip module, is being tested now. Like PAL-1, it will have military applications but also will have a huge effect on the world of commercial image processing.

"The first PAL-2 chips will be made this year," said Ritter.

So what does the new chip mean in terms of real work?

A huge image processing task, such as matching millions of finger prints in a national database -- a task that today takes an hour -- will take a 60 seconds with PAL-2. Airport security will be able to X-ray quickly and effectively every piece of luggage, not just carryons. Medical image processing -- mammography, telemedicine, cytology, robotic vision and even document image processing -- will move at lightning speed with greater clarity and accuracy.

The development of image algebra began at the University of Florida in 1984. Several image algebra programming languages have been created since then. The current version works with the PAL-1, an attached work station accelerator that has been used for sometime by the military.

Image algebra is a mathematical theory that simplifies and shortens lengthy computer commands previously written in long algorithms. When image algebra's more time-efficient commands are applied to the super-fast PAL-2 computer chip, the shortcuts will allow a computer to perform up to 10 billion operations per second.

Ritter compares the difference in calculation speed with what was realized when the world switched from ancient and cumbersome Roman numerals to the Indo-Arabic numbers we use today.

"Image algebra is an algebraic notation for specifying computer vision and image processing algorithms," said Ritter. "It is based on well-defined and well-known mathematical systems and grew out of a need by the U.S. Air Force Systems Command for a common image processing language."

Because the basis of image algebra is a mathematical theory that makes it independent of any future computer architecture or language, Ritter said, its longevity and continued evolution is assured. Also, he said, it can be programmed on any computer platform.

"Image algebra is living mathematics," said Ritter. "It is not a 'thing' but a body of knowledge, so it will continue to develop." -30-