Image understanding gets object-oriented
Tasks such as missile target seeking, pap-smear analysis, and checking for cracks in welds, which are either too dangerous, tiring, or demanding for humans, can be automated by computer. Now under a program sponsored by the US Defense Advanced Research Project Agency tools for this process of image understanding are being standardized in an image understanding environment (IUE). "The primary purpose of the IUE is to facilitate research and the exchange of research results," says Charles Kohl, director of research and development at Amerinex Applied Imaging (Amherst, MA), the company leading the development team.
As a standard implementation for models and algorithms, the IUE provides standard implementations of image-understanding algorithms that allow new techniques to be evaluated more quickly and help track progress in algorithm improvements. As such, the IUE is designed to provide a programming environment for rapid prototyping. "The program`s ultimate goal," says Kohl, "is to provide a software infrastructure of algorithms and tools for image understanding." A number of other prominent IUE organizations including Lehigh University (Bethlehem, PA), ADS/Booz-Allen (Mountain View, CA), Carnegie Mellon University (Pittsburg, PA), and Colorado State University (Fort Collins, CO) have participated in the project. So far, Amerinex has released an IUE core--software that consists of a graphical user interface, data-exchange tools, and control mechanisms for applying complex image-understanding algorithms. The IUE core is available from Amerinex via anonymous ftp (http://www.aai.com). With the release of the IUE libraries in early 1997, the company plans to supply a selection of algorithm libraries.
- ANDREW WILSON