Associate Professor Michael S. Brown, Department of Computer Science, National University of Singapore, talked in Canberra yesterday about "Interactive Visualization of Hyperspectral Images of Historical Documents".
Over time documents warp, fade, and the ink bleeds. As a result he decided to scan documents in 3D at the British Library. They then used an algorithm to "flatten" the document (much like map making). The 3D image can also be used for records of the preservation of the document.
At the Singapore Archives the problem was ink bleed. He pointed out that too complex an applciation will not be suitable for real world users. Instead they ask the user to show which bit of the document is the foreground and which is the background. The most appreciated feature for the users was that the original and corrected image could be displayed simultaneously.
Binarizationshop is a set of tools the project developed to allow the user to select the image algorithm to use.
He worked with "Art Innovation" who built a multi-spectral scanner for paper documents. The unit works by scanning using multiple passes using a single spectrum light source.
The idea is to collect more information, beyond the human visible range. Infra-red and ultraviolet light can be used to see what the human eye cannot. Also the measurements can be kept and compare over time to see how the document is aging. Essentially the document is treated in a similar way to the surface of the earth when analysed with a multi-spectral satellite. Just as a satellite image can be analysed to
detect a particular crop, the documents can be analysed to identify different inks, which look the same to the eye.
Unfortunately most of the results of this work are not freely available to the public. The result of the research work are in scientific journals which must be purchased and the documents which have been scanned are not made public by the institutions involved. It is a shame that so much publicly effort has been put into work which the public cannot see or use freely.