SoylentNews is people
SoylentNews is powered by your submissions, so send in your scoop. Only 15 submissions in the queue.
SoylentNews
AI turned a Rembrandt masterpiece into 5.6 terabytes of data:
A high-resolution image of Rembrandt's Nightwatch is now online. 717 gigapixels (yes, giga) to a claimed resolution of .0005-millimeters.
Last week the Rijksmuseum in Amsterdam posted an AI-constructed, ultra-high-res image of "The Night Watch" by Rembrandt. The original piece is nearly 15 feet long and more than 12 feet high and has been under intensive restoration since the early 1900s.
They've actually reconstructed some parts that had been destroyed over the ages, based on historical records.
Is a pixel size finer than the hairs on Rembrandt's brush enough detail for you?
Previously:
(2020-05-23) Revelations About Rembrandt's Masterpiece Captured on Camera
This discussion has been archived.
No new comments can be posted.
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
Best of all is never to have been born. Second best is to die soon.
(Score: 1, Insightful) by Anonymous Coward on Thursday January 20 2022, @09:45PM (7 children)
I'm trying to find a real paper describing the details behind this, but I don't care how many pixels they use to make up the image, they are certainly not resolving 0.0005 millimeters! That's 500 nm, or roughly the wavelength of green light. They took a bunch of pictures of it with a 100 MP camera and stitched them together. Your resolution is going to be whatever the resolving capability of the camera setup was (which is what I'm trying to find the details about). No matter what fancy image warping/remapping you're doing to make the panorama, you aren't increasing the resolution of what you're seeing. There are superresolution games you can play to claim resolution improvements, but not anywhere near orders of magnitude improvements.
(Score: 3, Informative) by hendrikboom on Thursday January 20 2022, @11:15PM (2 children)
The article and the Riksmuseum website disagree on the resolution. The article says .0005 mm; the museum says 0.005 mm.
I noticed that after submission, and couldn't correct it any more.
-- hendrik
(Score: 3, Touché) by hendrikboom on Thursday January 20 2022, @11:16PM (1 child)
And I just misspelled Rijksmuseum.
(Score: 2) by JoeMerchant on Friday January 21 2022, @06:07PM
Phonetic slip, Freudianly forgiveable.
🌻🌻🌻 [google.com]
(Score: 5, Informative) by ChrisMaple on Thursday January 20 2022, @11:58PM
Even 60 years ago, a top quality 35 mm camera's macro lens could resolve 5 microns, and microfiche lenses could do even better. Higher priced lenses and 60 years of progress have made much better lenses available. Getting down just 1 order of magnitude to the claimed 0.5 micron is not impossible or even terribly difficult if there's enough money to throw at it.
I'm not claiming they actually did achieve that resolution, only that it's possible.
(Score: 0) by Anonymous Coward on Friday January 21 2022, @01:41AM (2 children)
laser-guided five-axis camera positioning system
Hasselblad H6D 400 MS camera
CMOS, 100 megapixels (11600 × 8700 pixels, 4.6 × 4.6 μm)
53.4 x 40.0mm
4-shot Multi-shot mode
(Score: 0) by Anonymous Coward on Friday January 21 2022, @03:17AM (1 child)
What about the lens (fl and f-number)? Those are important details too.
(Score: 0) by Anonymous Coward on Friday January 21 2022, @03:14PM
https://cdn.hasselblad.com/f/77891/x/09de276a91/1601445-x-h-system-lenses-v3.pdf [hasselblad.com]