23
$\begingroup$

A few years ago, I read an article about a proposal Google put forward, as part of its paper-to-digital media project, which had the advantage of greatly speeding up the process of scanning text.

The idea was to scan the book using a high-frequency beam, which could be adjusted to allow for the thickness of each page. The average thickness of a book page is 0.009398 cm.

As I know nothing about any aspect of the instruments used in experimental physics, I wonder could anybody tell me if this is physically possible?

In other words, is there a X-ray range (which is my pure guess as to the frequency range involved), which would be suitable to scan a page, noting the contrast between the words and the paper and then move to the opposite side of the same page (or the next page and then turn the book over, if we cannot focus the beam finely enough)?

As far as I know, Google have not used this technology, so perhaps it's not feasible.

I'm not interested in any of the engineering involved (although I am happy to move it to EngineeringSE if required).

My question is based simply on my curiosity about the wavelengths needed to achieve this scanning, if it is in fact possible.

$\endgroup$
4
  • 4
    $\begingroup$ Probably the text on 2 adjacent pages would have to be scanned at the same time, because this text would be much closer than the thickness of a single page. Software would then have to 'resolve' 2 superimposed pages of text. $\endgroup$ Jul 23, 2016 at 14:21
  • 4
    $\begingroup$ The short answer to your question is "yes", it is possible AND probable depending upon cost/benefit at any moment in time. They could probably even employ some AI techniques to overcome any possible hurdles or even to augment the process. $\endgroup$ Jul 23, 2016 at 14:40
  • $\begingroup$ As Inquisitive says, this is really a trivial question with a trivial answer - Yes. At some time in the future, scanning a book in this way will be as simple as scanning a single page is today. But we do not yet know how or when. The only interesting (uncertain) aspect of your question is guessing how it will be done. $\endgroup$ Jul 23, 2016 at 19:14
  • 1
    $\begingroup$ The quantity "0.009398 cm" reads awkwardly. I would prefer to see "93.98 μm" instead $\endgroup$
    – Nayuki
    Jul 23, 2016 at 22:02

2 Answers 2

18
$\begingroup$

The proposal probably refers to terahertz (THz) imaging; THz is the band that is above microwave, but below infrared, between 100 um and 1 mm wavelength; often referred to as T-Rays.

Because of their relatively long wavelength the T-Ray penetrates well into non-conductive materials, and is far, far below the ionization threshold. Thus it is safe to work with, and can be used to see inside of stuff.

It's a relatively new field of application because it depends upon laser techniques for its generation; the frequencies are too high for electronic processes, but new techniques are always under development.

Terahertz and Cultural Heritage Science: Examination of Art and Archaeology goes into great detail, beginning with analysis of hidden layers in art, where it is used to explore underneath the visible surface:

enter image description here

In section 4.1 they give examples of scanning through multiple pages at the same time.

4.1. Historical Documents Often, historical documents on papyrus as well as on parchment and on paper cannot be read for a number of reasons. Sometimes the fragile sheets just cannot be separated because they are stuck together as a result of deterioration and damages. In other cases, the sheets have been reused as supports or covers for newer documents. There is considerable interest in reading this hidden information and a number of techniques have been tested to pursue this goal while preserving the documents. So far, X-ray computed tomography [66] has been the most successful method. Since THz-TD imaging was already evaluated for the inspection of postal envelopes [67], it was also tested with encouraging results for stacked papyrus layers written with carbon black ink [68]. Recently, even more successful results were obtained with a new and more sophisticated THz method called tomosynthesis [69] which was applied to image pencil writing on a stack of 50 paper sheets. THz tomography has been applied to resolve text on both sides of a single papyrus sheet [70,71]. THz was tested also for those cases where writing is obscured by stains and other inks in old purchament manuscripts [72] and it seems successful to characterize and evaluate conservation of iron gall inks [73] and parchment [74].

So with a properly calibrated T-Ray camera, and the tomographic software for image reconstruction, one can indeed scan through the pages of a book, top to bottom, never having to open the book - and without any hazard for the librarian. Just so long as the book isn't too thick.

Update: Demonstration of reading through nine pages - MIT News: Judging a book through its cover.

$\endgroup$
3
  • 2
    $\begingroup$ Report of research in [69] is promising but misleading. Although printed onto thin sheets of paper (post-it notes) the letters were each about 1cm square and they were on the 2nd, 25th and 50th sheets of a pad, ie about 3mm apart. They were not placed in vertical alignment but shifted so that they did not overlap at all vertically. Figs 8 & 9 show parts of each letter imaged with the others, despite Wiener Filtering. So this is far from equivalent to selectively imaging separate pages in a book. $\endgroup$ Jul 23, 2016 at 17:43
  • $\begingroup$ @sammygerbil: some of my colleagues work in this area - there will be improvements overtime. Better resolution is achieved with shorter wavelengths, but this requires more energy for greater penetration, and introduces more scatter. $\endgroup$ Jul 23, 2016 at 18:22
  • $\begingroup$ The wavelength will limit the resolution. To get 200 dpi (fax quality) scanning you need 8 ppmm, a spacing of 127 um. Hoping to do this with wavelengths longer than 100um seems likely to not work. The size of the test characters seems to validate this. It may be a useful technique for some applications but not book scanning. X-Rays might work if there is sufficient ink contrast but it will be a nightmarish de-convolution exercise to resolve the text from the mass of data. You will need to image at the desired pixel resolution with a 45 degree beam cone onto a 2D sensor. $\endgroup$
    – KalleMP
    Jul 24, 2016 at 21:27
15
$\begingroup$

Looking into this, the first thing that came up was a technology that used infrared light to determine the curvature of the pages of a book so it could be scanned non-destructively (in normal scanning, the pages need to be flat - before google's idea this was only possible with glass plates, which was inefficient, or dis-binding the book, which destroyed the book). This was patented with patent 7508978. A nice, concise explanation of this is available here. A diagram of this is shown below:

google book scanner

Now, for the x-ray, don't have to open the book scanner. As far as I can tell, this sounds a whole lot like multispectral imaging (more about this here). I couldn't find any references to google doing this, but it can be done using infrared light. This technology was used to read ancient documents where it appeared to be black ink on black paper, but infrared was able to distinguish (more about this here). This has also been used on paintings.

Another technology I found that uses x-rays was used to examine scraps of paper inside medieval bindings without taking off the binding (more about that here). This is called macro x-ray fluorescence spectrometry. More information about this is available here.

I'll be updating this as I find more information. Hope this helps!


Edit: I just today found another site about this that looks very promising. MIT has created a scanning system that can analyze the letters of nine pages. There is a video on the page that explains this in great detail.

$\endgroup$
2
  • 4
    $\begingroup$ I'm sceptical about the use of multi-spectral imaging because all of the hundred or so pages of text will have identical absorption/emission spectra. This is quite different from the problem of imaging ink through a layer of grime or plaster. $\endgroup$ Jul 23, 2016 at 14:31
  • $\begingroup$ @sammygerbil, I know. Looking into it, that was a option...I think macro x-ray fluorescence spectrometry might be the better solution. $\endgroup$ Jul 23, 2016 at 14:32