(06-11-2025, 10:14 PM)Bernd Wrote: You are not allowed to view links. Register or Login to view.isn't this a huge waste? 94% of the available tonal range just is not used
Scaling the pixel values from 0..2300 to 0.65535 would add no information and would only introduce noise.
Plus, if the cameras and lights were properly calibrated, the unscaled pixel values should let one compute the physical brightness (watts per square meter) and the actual albedo (fraction of light that is scattered). Then, if the pixels range 0..2000 for one page and 0..4000 in another, with the same lighting, it means that the vellum of the former is much darker overall than that of the second.
What
IS a big waste is using the inefficient TIFF format for those images. There are other image formats that can store the same data, without losing a single bit, in a much smaller file. (Even PNG gives a 50% reduction.) But TIFF has been the standard "full resolution" format for the last 40 years -- and no profession is more conservative and averse to change than Information Technology...
All the best, --stolfi
(06-11-2025, 10:14 PM)Bernd Wrote: You are not allowed to view links. Register or Login to view.Why not expose the image in a way that the histogram spans most of the range 0-65535?
I suppose that the extra bits would be mostly noise. The imaging element and its analog-to-digital converter probably give only 12 bits of data per pixel. Increasing the light level and/or exposure time would only overexpose most of the mage.
The Spectralon patch is indeed the brightest part of the image, except maybe in the 350 nm UV where its seems that it is only "light gray" and darker than other areas. And of course it is useless in the transmission images. (Those apparently lack a reliable "white" reference. There seems to be some translucent backing that makes even the wormholes non-"white". And also the lighting is uneven: brightest in the middle of the page, less bright along the left and right sides...)
All the best, --stolfi
(06-11-2025, 10:14 PM)Bernd Wrote: You are not allowed to view links. Register or Login to view.I assume whoever made these images knew what he was doing but one should never assume
Overall I agree, the images seem to be almost as good as one would want.
But I think there are some shortcomings in their protocol.
For instance, they (usually) took 3 images with each wavelength, with oblique light A, oblique light B, and both A and B at the same time. I think the latter is a waste of time, since the result should be exactly the sum of of the first two images. Even if there is non-trivial scattering through the vellum.
And both sources are tilted away by 30-40 degrees from the optical axis, so the image gets a strong texture because of the roughness of the vellum. Even in a combined A+B image, each tiny dimple will come out darker than its "true" color, and each tiny bump will come out brighter. This has been a big problem for my ink separation and un-painting experiments, since the variation in brightness due to surface roughness is much bigger than the hypothetical differences in ink or paint darkness...
And the two light sources are on the same plane with the optical axis (one tilted towards the top of the page, the other towards the bottom). By comparing those two images one can estimate the slope of the surface at each pixel -- but only in Y (the page's vertical direction). Instead they should have taken one image with nearly perpendicular light (to get the true colors), one tilted towards the top of the page, and one tilted towards the left side. Then one could estimate the average local surface slope in both X and Y directions. And then more reliably estimate the creases and curvature of the vellum.
(Perhaps they developed their protocol for paintings, where the background surface is generally smooth and the relief of the paint due to brush strokes is even more important than its color. Whereas for books on vellum the ink and paint are usually very flat but the background surface is rough...)
All the best, --stolfi
It's worth remembering that these images were taken not as part of a planned imaging campaign but as an impulsive decision while the team was at the Beinecke imaging the Martellus Map. That may explain some of Stolfi's observations. More here:
You are not allowed to view links.
Register or
Login to view.