Practical Systems

The network's performance in motion detection and edge detection means that it shows great potential for robot vision applications. Because the algorithms used in these applications have to be very complicated and require such large amounts of computing resources, this would seem an area in which integration onto hardware chips would dramatically improve the prospects for these concepts being taken up in practical systems [1].

The fact that the system can be used to achieve sophisticated image filtering relatively cheaply, and that the system works well on moving images, may make it useful for ultrasound applications. Ultrasound images are notoriously noisy, and it is usually difficult for the untrained eye to make out any useful features. The dramatic reduction of noise exhibited in the experimental results suggest that this would be excellent for making ultrasound displays clearer.

Finally, the fact that the system exhibits better performance at extracting large, consistent features from images shows great promise for the technology's integration into a hardware vision encoder, for encoding a continuous stream of images into an MPEG compressed stream. Although MPEG provides the potential for greatly improved transmission of video information, there is likely to be a great deal of consumer resistance to the fact that there is as yet no equivalent to the video recorders used for shooting home movies. A cheap real-time MPEG encoder will be needed to fill this gap in the market.