Apocalypto on TV

When the BBC heard about our efforts to Reveal the Unreadable they got in contact. They wanted to know if we had anything they could film and use for television. At the time we were working on the Bressingham Roll (see past posts) and were about to reveal the hidden text inside the roll.

A film crew came along for the day and the video below is the result.

Flying though the Bressingham Roll

To get some idea of the internal structure of the Bressingham Roll, it was instructive to generate a fly-though animation. This peels away the scroll, layer by layer showing what is underneath various layers of parchment and how they are stuck together (if they are).

The two videos below show two different “cuts” though the roll. The first is parallel to the long axis of the roll. The second is across the long axis of the roll – equivalent to making slices of a Swiss roll.

In this video, you can make out some writing – in places the ink was clear enough to give a good X-Ray contrast with minimal post-processing needed to make it visible.

You can see some dark marks on the parchment in this video, some of these are ink, others are some other material in the roll that is also showing up strongly in X-Rays. It’s much harder to recognise the writing in this view, because it isn’t presented to you in the usual format you see it.

Bressingham Roll – Initial Results

Soon after the Bressingham Scroll arrived, we performed a quick test CT scan to determine the settings needed so we could image both the parchment and any ink on it.

A single slice of the scan is shown here, the ink shows up as brighter patches on the whirls of the parchment.

The end-on single slice view of Bressingham roll shows both the parchment and evidence of ink very clearly.

This was imaged at 30kV with a 200um Al filter and post-processing with in-house developed beam hardening correction software.

Introducing the Bressingham Roll

As part of our continuing collaboration with the Norfolk Record Office, we’ve recently taken temporary possession of a parchment roll from a bundle in their archives (Call number PHI468,577X9)

Phi roll 002

The Bressingham Roll as unrolled as possible. The section to the right of the image is stuck together.

Phi 468 001

End view of the parchment roll in rolled up-state

The roll is unfit for production in the archives, with the inner portion becoming stuck together.

The ink used in the writing contains both iron and copper, so should have good contrast in X-Ray imaging.

Making Iron Gall Ink (Ferrogallic Ink)

As part of some outreach work, one of us (David) visited the conservation dept of the

Oak Galls to Ink

Westminster Archives, to assist with making some Iron Gall ink for use in the archives and to take some back to the Apocalypto Project lab for experimentation.

The history of iron gall ink manufacture includes plenty of alchemy, many recipes for ink include (some or all of)  urine, vinegar, blood, resin, wine and spirits. In the modern era we know that the active ingredients and a source of tannins and a source of iron ions.

We decided to standardize upon the following recipe.

Boil 2g of powdered oak galls in 200ml of water for one hour.
While solution is still hot, add 24g of Iron (II) Sulphate.
When all the Iron Sulphate is dissolved, add in 4g Gum Arabic.
Filter the solution and bottle.

Boiling oak galls for ink

Upon adding the Iron (II) Sulphate to the oak gall infusion, a dramatic colour change occurs; the pale orange/brown liquid turns deep blue-black, indicating that the tannins from the oak galls have formed chemical complexes with the iron ions. When all of the iron (II) sulphate dissolves the ink needs filtering to remove insoluble particles.


Filtering ink

The whole process took just over three hours, the majority of that time spent filtering the ink, finely powdered oak galls clog filter-papers wonderfully.

Meagen Smith has also written up the ink making adventure.

An Endorsement

This pleasing endorsement of the project arrived from Dr John Alban at the Norfolk Record Office.

Through my colleagues in the Norfolk Record Office’s conservation section, I have been following your project with great interest, and we were pleased to provide you with a ‘non-archival, non contextual off-cut’ on which to experiment. It has been intriguing to see the results of your processes for reading unrolled manuscripts and the way in which they have been improving. The latest example which I have seen has amazing legibility, and I am sure that this process is going to provide archivists and historians with an invaluable means of gaining access to the contents of rolled-up documents which, because of their condition, have hitherto been unreadable.

Yours sincerely,

Dr John Alban

[Posted with kind permission of Dr Alban]

Unrolling the Unrollable

Thanks to the efforts of Meagen Smith at Camberwell, The small scroll of parchment from the Norfolk Archives has been successfully unrolled. We can now compare the X-Ray digital unrolling with the until now hidden text.

Unrolled Norfolk Archives parchment

Digitally extracted text from the XMT scanned parchment scroll

It is clear that the extracted text is from the middle portion of the scroll – the text matches perfectly.

Revealing the Unreadable

New scroll, new challenges. First, we were pretty sure that our scroll is two-sides written,  same as the scrolls we processed before. The further analysis revealed that this parchment is one (OUTSIDE) side written, with lots of confusing bright spots of the metal dust on both sides of the parchment. Also we noticed the inside parchment surface generally is much brighter than the outside surface, with the intensity close to the ink intensity.

Slice from the top of the scroll with a tightly connected metal strip

The next challenge was the metal strip, which has the same intensity as the ink, and very tight connected with the parchment near the top edge of the scroll.  Although we can track the strip and remove it as a separate object in a similar way as we did for the two pages scroll, there are messy areas which impossible to separate without leaving parts of the strip on the parchment.




After the automatic unrolling the result looked promising, but not very readable:

After another week of experiments with the local segmentation, filtering, postprocessing and interpretation, almost everything become readable. The scroll appears to be an old, property related legal document.  As one can see on the original photos , the parchment is the cut fragment,  words are missing from both sides, left and right.


In December 2011 David and Graham visited the Norfolk Archives, based in Norwich, to discuss the Apocalypto project with Antoinette Curtis and her conservation staff. To assist  in our initial scanning and unrolling experiments, the Norfolk Archives graciously donated some non-archival, non-contextual off-cuts of parchment with hand written script using iron gall ink.

One of the items is a small, tightly wound roll of parchment, that would be very difficult to unroll to read or image by conventional means. This immediately attracted out attention, as it would be a real test of our ability to read an otherwise unreadable document.

Norfolk Archive mini scroll

The small, tightly wound scroll donated by the Norfolk Archives. It is long and thin and in a dessicated state that makes it impossible to unroll to read.

The scroll was scanned in our in-house MuCAT-2 microCT xray scanner with a 30 kV X-ray spectrum. Once the scan data was mathematically reconstructed, we could start to explore the scroll.

Norfolk Archives scroll x-ray cross-section, with an inset view of part of an inked letter.

The first thing we noticed in the x-ray cross-section images is the metallic foil strip glued to the inner edge of the parchment. This shows as the bright shape at the top of the scroll in the image to the left. We can also see indications of the iron gall ink, again these show as brighter areas along the line of the parchment surface. The inset image shows a view along part of a letter on the surface. The high curvature and short depth of field of the scanned slices means that only a portion of the letter can be seen.

Volume rendering the scan data and virtually removing the parchment reveals the metal components of the scroll

Applying volume rendering techniques to the scan data and virtually removing the parchment, we can explore just the ink and metallic components of the scroll.

The lime used in the manufacture of parchment also shows up well because of its calcium content – calcium being a metal. Excessive liming of the parchment can mask the visibility of ink in the x-ray scans; however in this case, while it shows up as a mass of spots and general ‘noise’ in the image, ink is still easily visible.

Adding back the parchment in the volume render of the scan data allows us to see the ink on the parchment, but due to the highly curved surface of the scroll, it’s impossible to read more than a few letters of a word using this technique.

The text on the rendered scroll looks faded and ratty. This is because the distribution of iron in the ink isn’t uniform, and it’s only the iron in the ink that provides contrast in the x-ray images.

For a full unrolling, the data was passed to Oksana and Paul, our colleagues at the School of Computer Science & Informatics at Cardiff University. They will take up the story from here.

A volume render of a section of the scan data, showing ink on the outer surface of the scroll.