Pictures of the Future    Fall 2006

Image Processing: Unlimited Horizons

Interview with Norbert Bauer

In which applications does digital image processing play a key role today?

Bauer: The world’s leading application area is automotive engineering, including everything from quality assurance to driver assistance systems. The automotive sector is increasingly demanding better and more intelligent solutions, such as cameras that record drivers’ blinking to determine if they’re tired. In such a case, the driver is alerted by steering wheel vibrations or other types of alarms. The processed food industry is another case in point. Here, digital image processing is being implemented in inspection systems that automatically detect impurities or check to ensure that packages are properly sealed, for example. In addition, there’s tremendous potential in the area of safety. Here, sensors can be used to monitor air quality and ambient temperature, or to film parts of burning buildings, and then send the data to a display mounted inside a firefighter’s helmet. Many advances have also been made in biometric facial recognition, which is now considered a reliable method of identification.

What are the goals of the Fraunhofer Allianz Vision?

Bauer: The alliance was established ten years ago to create synergies between the Fraunhofer institutes. To ensure that each part of the organization knew what the others were doing, we networked the image processing expertise of 13 institutes. Development redundancies have since been eliminated, and scientists are now working with shared development modules. This allows us to serve industrial partners like Siemens as a unified organization. Because image processing requires a great deal of interdisciplinary expertise, the alliance has proved quite successful. Basically, it allows us to assign the right people to the job at hand.

What trends have you identified?

Bauer: There are trends toward 3D imaging, texture analysis, high-speed cameras, color recognition and thermography. With today’s powerful computers you can implement image processing systems using relatively inexpensive hardware, and this has led to increasingly precise and detailed image recognition performance. For example, a laptop and a machine-based support system are now all it takes to analyze complex aerial and satellite images. Heat-flow thermography, which can be used to identify hidden weaknesses in windmill wheels and gas-turbine blades, is an up-and-coming area. I also see great opportunities for surveillance systems. In fact, automated image analysis could be refined to a degree where it would be possible to identify pickpockets in large crowds on the basis of their movements.

What else will be possible in the future?

Bauer: At the moment, soccer-playing robots offer the best example of what we can expect to see. They can interpret the scene in front of them, precisely measure distances, and coordinate their movements. And they do all this very rapidly and with practically no errors. Still, we can expect to see further advances in processing speed and the analysis of complex situations—for example, in the area of camera-based assistance systems in cars.

What will image processing be like in ten years?

Bauer: I think we’ll see the greatest amount of innovation in data processing, recognition software and optical resolution performance. One example involves photonic mixer devices for 3D imaging. These sensors estimate distances very quickly by comparing transmitted beams of light and their reflections. Further momentum will come from self-learning algorithms that describe not only individual pixels but also complex relationships. Unfortunately, not enough scientists are conducting research in this area today. Other new applications will include the monitoring of large crowds and the analysis of movement patterns and facial expressions as a means of gauging crowd moods. Personal identification based on 3D recognition of facial and bodily features will become more important, as will the surveillance of airspace, borders, and traffic and transport systems.

Are there still some areas where problems need to be worked out?

Bauer:Yes, in terms of details. For example, people change. They can cut their hair or grow a beard, thereby creating a difficult situation for a system that works with reference images and image comparisons. The same applies to color images in rapidly rotating printing machines, where a camera must determine within milliseconds whether color structures conform to quality guidelines. The effectiveness of such applications is still limited today—but that will change quickly in the future.

                                                                                     Interview by Andreas Beuthner