If you've ever copied pages from a book, you're familiar with the problem - dark, distorted
words where the page is bound into the book.
To correct it, most of us use the "brute force"
method for getting readable copies - pushing the book to flatten it against the glass scanning
surface, called the platen.
XIG researchers think there is a better way.
They have proposed a
simple and elegant solution - a mathematical formula that could be incorporated in the software
of common scanners to eliminate the problem. It could not only save the spines of books but
also produce copies that are easier to read.
The innovative research was presented by Beilei Xu, a scientist at the company's Webster, N.Y.,
laboratory, at the 5th International Conference on Imaging Science and Hard Copy in Xi'an,
China.
According to Xu and Robert P. Loce, another Xerox researcher who worked on the project, when a
book page is not in uniform, intimate contact with the scanning surface there are actually two
distinct problems.
The variation in illumination causes some portions of the copy to be darker
than others, and the variation in distance from the scanning surface causes letters or objects
farther from the surface to look warped.
There have been some specialized and costly attempts to correct for this.
At one time, for
instance, Xerox sold a copier with an angled edge and articulated cover so people could copy
pages without cracking books all the way open.
And another solution is dedicated book scanners
with height sensors, so the book lies face up, and scanning takes place from above it.
Instead of redesigning the machine to accommodate books, however, Xu and Loce decided to build
the capability into the software.
Scanners capture images by shining light onto the image; that
light is reflected off the image, then measured by sensors, which "read" the lightness or darkness
of the image, pixel by pixel.
"Normally the light only provides information on the reflectance of the original document,"
Xu said.
"In our application, we use the sensed light to also determine the distance of the
book from the platen for each pixel on the page."
Because parts of the page farther from the platen appear darker, they can determine the
variation in distance and mathematically compensate for it, eliminating from the image the darker
portion where the page is bound into the book.
However, the image will still have distorted
words at the center.
The second step in their process is to apply a de-warping factor, again
based on how far the page is from the platen, to compensate for the geometric distortion of the
image.
The patented process could readily enable books to be scanned with high-quality results on
almost any inexpensive, conventional scanner, according to Loce.
The software is designed to
work with the methodology that is commonly used in the consumer market scanners and copiers, an
area where Xerox does not currently compete.
However, the technology is available for license,
and the theory behind it could be adapted in the future for the high-speed scanners used in Xerox
products.
A video of this technology is available here.
|
