- Options
- Login
- Desktop Style
Options
f17r
- f17r
-
RE: f17r
Jorge_Stolfi > 12-11-2025, 08:57 PM
(12-11-2025, 07:32 PM)oshfdk Wrote: You are not allowed to view links. Register or Login to view.But is it possible to determine if they became transparent or just nearly the same shade as the parchment?
I am not sure I understand the question, or what difference it would make. Here is the histogram of pixel values in the MB940IR_012_F image of f17r, clipped to the page proper minus a narrow margin all around:
Note that the histogram count scale is lograithmic, and that of the cumulative histograms is "bilogarithmic", with log-like behavior near 0 and near 1.
Here is the histogram of pixel value over the six patches of the gray scale that was placed just above the page, near the upper right corner of the image:
And here is the histogram of the Spectralon patch, that was placed just above the page, near top center of the image:
Spectralon is a white material used as a standard in technical imaging. It is advertised as scattering 99% or something of the light (that is, being "white") at all wavelengths except UV, where its reflectance drops. So you can compare brightness levels in images taken at (say) 700 and 840 nm by scaling all pixels in each image so that the Spectralon patch has the same gray shade in both images.
The Spectralon histogram is not a sharp peak because it has a mottled appearance:
The first image is just contrast-stretched (linear map), the second one is both contrast- and brightness-stretched. I don't know whether that mottling is normal, or the result of dirt having stuck to the surface of the patch. Anyway, as long as the dirt is the same on all images, the peak of the histogram for each image can be used as the "white" value. I myself need a table of those values and other parameters, but I haven't made it yet.
For this image, specifically, a perfectly white matte surface should have pixel values a little above 2160. (Assuming the dirt is matte too. The light direction is about 45 degrees, so those brightest spots should not be shiny-finish highlights...)
The last hump in the full-page histogram is the vellum. It is not a sharp peak because of warps and stains of all kinds. It is obviously well below the Spectralon white.
The gray patches are not as accurate as the Spectralon patch and their reflectance varies a lot with the wavelength outside the visible range. So they should not be used to calibrate the pixel scale of the IR images.
Here are the coordinates of the various parts of the TIFF images (in the ImageMagick's "-crop" notation). The "whref" is the Spectralon patch.
Code:# To be sourced by shell scripts. The {*_crop} parameters are relative
# to the original image, before rotation.
page_crop='6500x4650+0934+1352' # Full page with a bit extra margin ("full.png").
txref_crop='0196x4650+7405+1352' # Transmission ref patch ("txref.png"); same Y range as page.
hist_crop='6050x4000+1100+1584' # Region to consider in "full" histogram.
C0_crop='0400x0300+1488+2008' # "C0" image clip (test).
C1_crop='0320x3700+0945+1952' # "C1" image clip (top marginalia).
C2_crop='0850x1800+2712+2500' # "C2" image clip (lines 7-11, end).
C3_crop='1320x0900+1400+4660' # "C3" image clip (lines 1-6, start).
scale_crop='2200x0200+5822+0688' # Coords of gray scale patcthes ("scale.png").
whref_crop='0300x0300+3422+0418' # Coords of white ref patch ("whref.png").
grref_crop='0300x0300+4162+0406' # Coords of gray ref patch ("grref.png").
ruler_crop='2400x0170+5710+0916' # Coords of ruler ("ruler.png").
rot=90 # CW rotation to apply (degrees)
Hope it helps, --stolfi -
RE: f17r
oshfdk > 12-11-2025, 09:07 PM
(12-11-2025, 08:57 PM)Jorge_Stolfi Wrote: You are not allowed to view links. Register or Login to view.I am not sure I understand the question, or what difference it would make.
Writing in a black sharpie on black paper would be largely invisible, this doesn't mean that the sharpie became transparent. I was suggesting that maybe in IR the parchment itself has about the same shade as the ink, which would make the ink blend in with the parchment.
However, looking at the page the ink is nearly invisible not only over the blank vellum, but also over the plant image see through from the reverse of the page, so I think it is indeed quite transparent in IR. I don't know what this means in practice, but if this is inconsistent with iron gall ink, this would be interesting. -
RE: f17r
Jorge_Stolfi > 12-11-2025, 10:52 PM
(12-11-2025, 09:07 PM)oshfdk Wrote: You are not allowed to view links. Register or Login to view.I don't know what this means in practice, but if this is inconsistent with iron gall ink, this would be interesting.
Indeed. But I would rather understand it better before I discuss it.
Meanwhile, here are the images of the starts of lines 1-7 under 625 nm (red) and 700, 735, 780, 870, and 940 nm (IR), reduced in size 50% and separately normalized to 0-1 range (thus not in the same color scale):
And here is approximately the same area in the BL 2014 image:
So the fading of the text as the wavelength increases is gradual, which seems compatible with "ocher" ink but not iron-gall ink. But it seems also that the glyphs and parts of glyphs that are noticeably darker in the BL 2014 image (which my Superior Pareidolia says are retraced at a later date than the rest) keep their opacity better, up to 940 nm. While that may mean a different "ocher" composition (with more MnO2?), it may also mean that the ink of those strokes contains some iron-gall ink.
All the best, --stolfi -
RE: f17r
Bernd > 12-11-2025, 11:00 PM
Spectralon is known to attract dirt due to its porosity. Some mottling is normal but the dark spots are clearly dirt. I don't think this has any significant impact though.
I agree that it is highly unlikely the ink consistently has exactly the same reflective properties as vellum (+underlying paint) in IR. Therefore I think we can safely assume the ink becomes transparent in IR.
I think trying to replicate the spectral properties of the ink would make a nice master's thesis. I wonder why this was never done considering the MSI images are over 10 years old. At least testing a few common contemporary inks for their properties in UV and IR shouldn't be too hard. -
RE: f17r
Bernd > 12-11-2025, 11:21 PM
Regarding the noticeably darker single letters - can we determine whether this is another layer of different ink that was applied later or just a thicker layer of the original ink due to irregular ink flow from the quill pen? I must say those isolated darker letters which regularly pop up are curious but we must also take into account that this was not done with a modern instrument that provides uniform ink flow.
So I wonder - do the darker letters behave differently under the UV-IR spectrum which would point to ink with different composition - or is the darkness just the result of a higher thickness of the ink layer with identical composition than the rest of the ink?
Again, this is something that should be put to the test with 'experimental archaeology'
I doubt we could determine this from the MSI images though, it would probably take a real spectrometer that gives us more reliable data than a photograph. RAMAN would be awesome to tell us a lot more about the ink and paint. -
RE: f17r
oshfdk > 12-11-2025, 11:27 PM
(12-11-2025, 11:13 PM)nablator Wrote: You are not allowed to view links. Register or Login to view.This seems consistent with iron-gall ink: You are not allowed to view links. Register or Login to view.
I'm not sure, it looks like there the ink is still quite visible in IR+ (1000nm), and only becomes completely transparent in IRR (~1500nm?). With MSIs of the Voynich MS it's barely visible below 1000nm already.
I misread the chart, it's nearly transparent in IR+. -
RE: f17r
Jorge_Stolfi > 13-11-2025, 12:31 AM
(12-11-2025, 11:13 PM)nablator Wrote: You are not allowed to view links. Register or Login to view.This seems consistent with iron-gall ink: You are not allowed to view links. Register or Login to view.
According to that graph, at 940 nm iron-gall ink should still have less than 40% reflectivity. So yes, maybe the darker glyphs in that BL2014 clips are iron-gall ink.
But the color is still off, though. I would expect it to be black...
Maybe whoever retraced those glyphs tried to match the old ink's color by mixing some ocher watercolor paint (brownish yellow-orange) and iron-gall ink (black) to get a similar shade of dark brown?
All the best, --stolfi -
RE: f17r
JoJo_Jost > 17-11-2025, 11:02 AM
Excuse me, but I would like to say something else.
Of course, it may be that a retracer thought he had a Latin sentence in front of him and, like us here, tried to find a solution. So he found a solution that seemed plausible to him and then ‘wrote’ this Latin sentence into the Voynich glyphs (which would explain a lot, the superscript P's and the Voynich T further back and other things).
But that doesn't interest me at all in this case. I was only interested in:
IF it is Latin, what is the most coherent translation? And that seems to have been misunderstood.
I came up with the sentence because I did exactly what one does with illegible passages, especially in old handwritten manuscripts
You look at the glyphs, divide them into those with ascenders and descenders, and insert variables, so to speak: the "l" could be an "l" or an s or an f, perhaps you can't see the belly, then you're left with b, d, k, p. Very simple mathematics. Variables, then check for coherence with the word and pay attention to the context.
The only thing I did differently was that I did it a little more consistently. Why wasn't "f" considered a possibility before? No idea.
Luc3: Since L has not yielded any results so far (9 years)
= Xuc3
I tested the usual suspects with X, including f (the f with a loop is verifiable).
Then I checked fuc3 as a possibility. I wasn't sure about the c, so I created a new variable:
= fuX3.
I ran through that and found, among other things: Furus (thieving) funus (corpse) fumus (smoke) futurus (future) fucus and also fusum. If you check that in context, Fumus (smoke) might have worked, but the m as a glyph is too far away. So that left Fucus / fusum.
And I proceeded in exactly the same way with ‘mulhier’. The h couldn't be an h, see 4th word, first letter. But then nothing came of it, so I used variables again. In the context of Fucus (very rarely propolis (bee resin) and fusum, Apiarius was then more likely again. And so on.
Conclusion: No eisegesis, just pure exegesis and logic.
And whether you like it or not:
If the sentence on 17r is Latin, the following translation is one of the most coherent translations you will find in the context of a recipe book.
m. Apiar' appar' fuc3 her'
Have a good time. -
RE: f17r
Jorge_Stolfi > 17-11-2025, 01:46 PM
(17-11-2025, 11:02 AM)JoJo_Jost Wrote: You are not allowed to view links. Register or Login to view.thought he had a Latin sentence in front of him and, like us here, tried to find a solution
That is not my proposal. I think the Restorer of those lines was not trying to recover the sentence, or even the words. His primary intention was just to restore the characters. And he may not even have assumed that they were Latin letters; he just "saw" Latin letters in the faint smudges that remained because those were the characters he was familiar with. (Why should I be the only person in all history with Superior Pareidolia?)
It is even possible that he (unlike Barschius) had not spent much time looking at the book, and thus had not yet learned the Voynichese alphabet...
(17-11-2025, 11:02 AM)JoJo_Jost Wrote: You are not allowed to view links. Register or Login to view.IF it is Latin, what is the most coherent translation?
I believe that the "restored" text made no sense even to the Restorer himself. Thus I would say that this line of inquiry is a waste of time. Which it seems to have been for the least 100 years...
All the best, --stolfi