Though they are some of the most-sophisticated pieces of equipment in the office, copy machines don't possess a whole lot of glitz—a perception worsened by the sight of a repairman next to a heap of copier parts. And, now that photorealistic computer printers are everywhere, even the slightest image flaws produced by a copier are noticed. In response, researchers at Xerox (Webster, NY) are working on a diagnostic technique that could allow copier image flaws to be quickly recognized and fixed, with the repairman relegated to a secondary role.
Called advanced image-quality diagnostics (AIQD), the technique relies on a personal computer (PC), a scanner, and specialized image-processing software. Following instructions given by a PC program, the customer scans a degraded image from the black-and-white or color copier under test into the PC (the scanner must be prequalified by Xerox for AIQD use). The software analyzes the image and determines what is wrong with the copier. Then, AIQD issues instructions for copier repair that is to be done by the machine itself, by the customer, or by a technician.
The customer is given a set of predetermined test patterns; each individual pattern is used for a certain subset of all the possible defects, according to Meera Sampath, a principal scientist at Xerox. The AIQD system directs the user as to which specific test pattern or patterns of the given set will need to be copied during any run of the system. Image defects that the software can detect include lines/streaks/bands, uniformity variations, spots, mottle, deletions, light or dark images, and incorrect color.
"The software for detection of defects is based on human visual-perception modeling and image-processing techniques," says Sampath.1 "The visual-perception modeling helps to mimic the human eye and get a quantitative measure of objectionability of an image or a defect in a image. Image-processing techniques such as Fourier transforms and others are used to characterize the defect in quantitative terms—for example, what is the frequency of the band, what is the amplitude of the band, and so on. The quantitative descriptions obtained using the image-processing techniques may further be translated into qualitative terms using feature-extraction techniques. A few typical features are 'light streak,' 'streaks separated by x cm,' and 'streaks every alternate page.' "
From the scanned images, the software also determines what must be fixed in the copier. The AIQD system has a diagnostic engine based on probabilistic reasoning that decides what the copier's problem is, says Sampath. The input data can be obtained from the scanned image, or they can be machine data obtained from running diagnostic tests in the copier. The latter include damaged cleaner blades, worn fuser fingers, scratches in the photoreceptor, contaminated or worn donor rolls, dirt, and so on.
"Today, many of our copiers and printers are built around the notion of customer-replaceable units," says Sampath. For this reason, AIQD is designed to direct the customer to exchange modules if necessary. Unlike the multitude of metal parts the proverbial copier repairman strews about, the modules are easily swapped.
"Automated defect-compensation schemes are currently under research or are being planned for future product releases," says Sampath. The ultimate result—a self-healing copier—would be a star in the high-tech office-equipment world.
REFERENCE
- D. Rene Rasmussen et al., Proc. IS&T's 1998 PICS Conf., 174 (Portland, Oregon; 1998).