Depositphotos
Alex Kachkin, a researcher at the Department of Mechanical Engineering at the Massachusetts Institute of Technology (MIT) in the United States, has developed a method that allows restoring old damaged paintings in a matter of hours with the help of artificial intelligence.
In his spare time, the scientist restores paintings as a hobby. Kachkin learned that many galleries have many paintings that are never exhibited because of their poor condition and the need for complex restoration. Restoring these paintings by hand would be very expensive and take a lot of time.
However, the researcher developed the method allows you to to restore old damaged paintings not for years or months, but in just a few hours. Kachkin tested his method on an old damaged 15th-century oil painting. Using traditional cleaning methods, he removed excess paint applied during previous restorations and performed a high-resolution scan of the painting.
After that, Alex used artificial intelligence algorithms to analyze the scanned painting, identifying more than 5 thousand damaged areas. The scientist then created a digital copy of the painting in its original form. He also used AI to create a map of the painting with damaged areas, cracks, and places where there is no more paint. This map also showed the exact colors that needed to be applied to these areas to restore the original appearance of the painting.
The scientist used a high-quality inkjet printer to turn the digital map of the painting into a two-layer mask printed on an ultra-thin, transparent polymer film. One layer contained the colors that were supposed to replace the damaged areas in the appropriate places, and the other layer was white.
«To fully reproduce a color, you need both white and colored ink to get the full spectrum,» Kachkin explains
At the final stage of the restoration, the scientist leveled the mask and applied it to the surface of the painting, securing it with a thin layer of varnish, which he applied by spraying. Both the mask and the varnish can be dissolved without damaging the original paint on the painting.
For the project, Alex used a total of 57,314 different colors to restore 5,612 separate areas of the painting, and it only took him about three and a half hours. He calculated that doing the same task manually would have taken 66 times longer.
The results of the study were published in the journal Nature
Source: New Atlas