How do you think the new GigE standards will influence the machine vision industry?
Respond or ask your question now!
By Hank Russell
Back in the 1980s, the only big image compression standard was JPEG. A decade later, JPEG2000 began to gain some name recognition. "JPEG is very widely accepted," says Jack Berlin, president of Pegasus Imaging Corp. (Tampa, FL). "It's been around a long time. It's in your browser, everybody handles it and their digital camera takes it. JPEG2000 is less accepted, but it's becoming more accepted now."
Although these two international standards are easily recognized, they are not the answer in all applications. Space organizations who perform remote sensing utilize JPEG-LS, a lossless/near-lossless compression standard developed by HP Labs (Palo Alto, CA) based on the LOCO-I (LOw COmplexity LOssless COmpression for Images) algorithm, for use of continuous-tone images. Graphic artists who design websites use the GIF (Graphic Interface Format), a low-color lossless image compressor based on the LZW (Lempel-Ziv-Welch) algorithm. For video compression for broadcast applications, the Motion JPEG (M-JPEG) or the latest MPEG standards work.
Image quality is always a concern when compression comes into play. "The issue of image quality is, of course, a trade-off between the quality that is wanted and the bandwidth, or space, that is available for storage," says David Johnson, product manager for Curtiss-Wright Controls Embedded Computing's Cambridge, UK, office. "In a real application, what might happen is a certain bandwidth is available, or a certain target space is available, and the aim is to achieve a certain level of compression. That implies a certain loss of quality, and it's just getting that trade-off right between how much loss the user is prepared to accept against how much bandwidth or storage space they would want to throw at it."
The most important thing for image compression, emphasizes Berlin, is that the standards be used. "People that don't use standards in software are setting themselves up for trouble someday," he says. "People that have compressed into non-international standards formats not only haven't gotten good compression as advertised, but they find themselves with images they can't do anything with because whoever can uncompress them is now gone."
Johnson adds that another issue is latency, especially for military imaging applications. "The key is the man-in-the-loop operation," he says. "In other words, if somebody is looking at what they could see on the screen, and then doing something on the basis of that -- maybe moving a joystick or a mouse so they're trying to align something, or point a gun -- then that's where latency is critical because, if you put latency back into a feedback control loop, it can become unstable."