Replicas of famous paintings are routinely created with printers that use only four inks – cyan, magenta, yellow, and black. RePaint, a new technique developed at MIT, combines artificial intelligence, 3D printing, and a rich 10-ink palette for much more faithful results in any lighting condition.
Pictures of the original works of art are first analyzed by a deep-learning artificial intelligence algorithm, which selects the right combination of inks to be used. Each of the 10 transparent inks is then carefully deposited by a 3D printer, with colors often being stacked on top of one other. Gradients are reproduced through well-known techniques such as half toning (the use of dots of various size and spacing to simulate a continuous tone).
The result, according to the researchers, is a color accuracy that was measured as four times greater than state-of-the-art physical models and that, unlike standard prints, is faithful to the original irrespective of placement or room lighting.
"If you just reproduce the color of a painting as it looks in the gallery, it might look different in your home," says Changil Kim, one author of a paper on the subject that will be presented next month. "Our system works under any lighting condition, which shows a far greater color reproduction capability than almost any other previous work."
Besides commercial prints of famous paintings for your home, the RePaint technique could be used in museums to create accurate replicas of works of art which have been stolen or are too sensitive to be exposed to the public.
However, as of now, there are still a few kinks that need to be worked out. The current palette is incomplete, lacking colors such as cobalt blue, and work is still in progress to produce glossy and matte finishes. Lastly, the printing process is quite slow, with the prints being only about the size of a business card, although the researchers feel optimistic that future commercial 3D printers will speed up the process considerably.
You can watch the algorithm at work in the video below.
Source:MIT
