Slashdot Mirror
How Google's High Speed Book Scanner De-Warps Pages
Hugh Pickens writes "Patent 7,508,978, awarded to Google, shows how the company has already managed to scan more than 7 million books. Google's system uses two cameras and infrared light to automatically correct for the curvature of pages in a book. By constructing a 3D model of each page and then 'de-warping' it afterward, Google can present flat-looking pages online without having to slice books up or mash them onto a flatbed scanner. Stephen Shankland writes that the 'sophistication of the technology illustrates that would-be competitors who want to feature their own digitized libraries won't have a trivial time catching up to Google.' First, a book is placed on a flat surface, while above it, an infrared projector displays a special mazelike pattern onto the pages. Next, two infrared cameras photograph the infrared pattern from different perspectives. 'The images can be stereoscopically combined, using known stereoscopic techniques, to obtain a three-dimensional mapping of the pattern,' according to the patent. 'The pattern falls on the surface of (the) book, causing the three-dimensional mapping of the pattern to correspond to the three-dimensional surface of the page of the book.'"
Obviously it was worthy enough to be issued; but I don't know how worthy it is in the broader sense.
Notably, for instance, there has been a fair bit of interest, for some years, in using digital cameras in concert with projectors, either for automatic keystone/distortion correction, for projectors that aren't perfectly aligned with the projection surface, or for automatic coordination of multiple projectors illuminating the same surface, without laborious manual tiling adjustment. This is, in essence, an equivalent problem(inferring a surface's geometry based on pictures of a known image projected upon it).
The IEEE has held "Projector-Camera systems" workshops since 2003, and somebody was obviously working on it before that. I'm not saying that Google's patent falls into asshole troll territory or anything; but the notion of doing surface geometry inference based on known image projection isn't nearly as novel as it might seem.
I believe the pattern barcode scanners use is simply trying to look for the barcode in several different directions, but I could be wrong.
I also believe there's either rudimentary correction for common types of distortion (i.e. on cylindrical objects) or just wide enough tolerances to allow it to work anyways.
Of course nobody reads the FAQ! If people read the FAQ, the Questions wouldn't be so Frequently Asked.
Google is mostly scanning books borrowed from university libraries. Librarians get cranky if you borrow a book and return a stack of loose sheets of paper.
Search 2010 Gen Con events
You jest, but this technique *has* been around for years. I remember when digital cameras first became available there was a product that could perform a 3D scan by projecting a pattern onto the object and using an offset picture. I think the pattern came on a slide - that's how long ago it was! Here's a whole wikipedia page about the scanning technique: http://en.wikipedia.org/wiki/Structured_Light_3D_Scanner
This picture is especially good: http://en.wikipedia.org/wiki/File:6-seat.jpg
Anyway after reading the patent abstract, it isn't about the 3D scanning at all, it appears to be about an algorithm to find the fold once you've already got the point cloud. I would have thought that was fairly trivial. A possible approach would be to take the radon transform of the height map and find the smallest value that's roughly in the middle.
It's simply called adaptive optics (AO). In AO, a guidestar is a natural isolated point-like star that is close to your science object (what you are trying to look at). If a laser is used to excite the sodium layer to create an artificial reference, it's called a "laser guidestar".
Anyway, this "trick" is completely different from adaptive optics in both the mathematics and implementation.