+- The Voynich Ninja (https://www.voynich.ninja)
+-- Forum: Voynich Research (https://www.voynich.ninja/forum-27.html)
+--- Forum: News (https://www.voynich.ninja/forum-25.html)
+--- Thread: [Blog Post] MSI at last! (/thread-4363.html)
RE: MSI at last! - oshfdk - 17-09-2024
(Please see You are not allowed to view links. Register or Login to view. for an update including the code you can run to reproduce this.)
It should be possible to reproduce the top right image quite easily, this was just a linear regression.
Take the 38 TIFFs for 1r sorted alphabetically and multiply their pixel values using the following coefficients (one coefficient per image):
[ 2.86566222e-03, -7.31746714e-04, -7.91509891e-04, 6.26958368e-04,
-5.44016711e-04, 2.13091826e-04, 6.41224414e-05, -1.15745158e-03,
-1.19574426e-03, -1.41510511e-03, -1.77602363e-04, -1.28772144e-03,
-2.78448038e-04, -2.21850204e-04, -5.14871113e-04, 1.41009356e-03,
1.19763900e-03, 5.67961854e-04, 1.06700996e-03, 2.66545916e-04,
1.35468789e-03, -5.18381671e-04, 1.60250474e-04, -6.39294269e-04,
-6.06875071e-04, -5.26037804e-04, -8.50949681e-04, -1.46222198e-04,
2.00271206e-04, -4.03289436e-03, -6.48766122e-04, -6.49732137e-05,
-3.30974503e-03, 3.10579239e-03, 7.33322792e-03, 3.67331676e-03,
3.36502464e-03, 1.67928168e-04]
and then sum the resulting images together, add bias of 1.6710087437575252, multiply by 256 and clip all values outside [0, 255] range, the result should be an 8-bit image.
RE: MSI at last! - oshfdk - 17-09-2024
(17-09-2024, 01:26 AM)ReneZ Wrote: You are not allowed to view links. Register or Login to view.Another thing (back to f1r):
Brumbaugh writes in his book (p.115-116):
"After some work with ultraviolet light on folio 1r, my son Robert Conrad Brumbuagh, who was assisting me, proved that indeed someone had recognized the Bacon attribution cipher and had written it here in the margin. It is now badly obliterated and faded. In very small numbers, just above this table, is a date 1 * 3 0, the * illegible. Now, 1630 would fall into the period between the death of Tepenecz (in 1622) and Marchi's inheriting the manuscript (prior to 1644). At first, I thought the table might be earlier, a deliberate invitation to any would-be purchaser to read the ''Bacon'' cipher in the key. But we now favor the 1630 date, and assume that the writer of this table, having found that it read the key text, hoped that it would work for the balance of the cipher manuscript text as well."
There is quite a bit to unravel here, but the one thing that I have always found interesting is his reference to these tiny numbers above the cipher tables. I have looked and never seen anything.
Can we see this now?
(Note: Brumbaugh also thought that below the Tepenec signature was the word "Prag" but this we now now says "No 19".)
The models don't detect anything there. For the convenience I'm adding links to the cropped part of the upper right corner of 1r image taken from each of the TIFFs separately. There is minimal color processing, I just set the bottom 1% of the brightness to black and the top 1% to white and scaled the rest proportionally (otherwise most of these images are too dark to view). There are three versions with gamma set to 1, 2 and 3. I can't see anything resembling a year anywhere.
Gamma1 (dark and bright parts are equally represented): You are not allowed to view links. Register or Login to view.
Gamma2 (more details in the dark parts of the image): You are not allowed to view links. Register or Login to view.
Gamma3 (even more details in the dark parts of the image): You are not allowed to view links. Register or Login to view.
RE: MSI at last! - RobGea - 17-09-2024
To speculate on the invisible thing on f116v, kinda reminds me of a seal or stamp.
RE: MSI at last! - Koen G - 17-09-2024
I find it hard to say what's going on there. Some of the picture looks a lot like the regular noise on the page (marked green). What makes this special is probably the very straight lines, marked in blue on the right. Might these be smaller versions of the folds that are visible on the rest of the page? This is quite a battered piece of vellum.
RE: MSI at last! - oshfdk - 17-09-2024
(17-09-2024, 12:40 PM)Koen G Wrote: You are not allowed to view links. Register or Login to view.I find it hard to say what's going on there. Some of the picture looks a lot like the regular noise on the page (marked green). What makes this special is probably the very straight lines, marked in blue on the right. Might these be smaller versions of the folds that are visible on the rest of the page? This is quite a battered piece of vellum.
It is quite possible for this to be a mix of different features (folds and generic noise). However, I think the vellum doesn't show any visible folds in this area.
This is the overlay and the clean scan side by side.
RE: MSI at last! - oshfdk - 17-09-2024
No visible folds on the other side.
RE: MSI at last! - oshfdk - 17-09-2024
There are several ways to fill in the missing parts, if we treat this as a signature (I have no idea whether it is a signature or any kind of writing at all, to be clear).
I tried writing it this way, doesn't look that bad
I know nothing about medieval signatures, but I was under impression they used to be much more elaborate? This doesn't have a medieval vibe to me.
(Just to avoid any possibility of wrong interpretation, the image in this post is my attempt to recreate a possible signature-like form based on what I subjectively see in the processed MSI images, using a whiteboard marker, this image is not from VMS, this is just a photo of my whiteboard nothing more
)
RE: MSI at last! - oshfdk - 17-09-2024
(17-09-2024, 07:11 AM)oshfdk Wrote: You are not allowed to view links. Register or Login to view.It should be possible to reproduce the top left image quite easily, this was just a linear regression.
Here is a simple Python script that takes the original MSI TIFFs for 116v (you can modify it for any other page, as long as the number and order of the files is the same) and produces a combined image with clearly seen feature in the lower right corner of f116v. For simplicity the script assumes all TIFFs are in the current folder, you can change the path at the top of the file if needed.
You are not allowed to view links. Register or Login to view.
That's the tiny version of the result you should get (the original is approx. 26 megabytes PNG, 8176x6132, greyscale 8-bit):
The original TIFF files (about 5GB for f116v) courtesy of Lisa Fagin Davis are available at:
You are not allowed to view links. Register or Login to view.
RE: MSI at last! - oshfdk - 17-09-2024
I got a question in DMs about how exactly I trained for the missing ink, since the answer is very simple, I'll copy it here, I guess this might be interesting for more people:
It's very simple really, I think I mentioned that in the thread: I just used for the positive points about 50/50 the points with visible ink and the points where ink should have been, but is not visible. E.g., you have an EVA t, but small parts of its loops or legs are absent (faded, the ink didn't take, etc), you still know perfectly well where they should have been, so you dedicate about 50% of positive training points to these spots. That's the whole secret recipe.
RE: MSI at last! - Koen G - 17-09-2024
This really brings out the text very well.