Although AI-based restoration methods can indeed bring new life to damaged paintings, the end result is typically a digital copy of the original painting. By contrast, a new MIT technique applies reversible repairs to the physical painting itself, in the form of a removable mask.
The process was developed by mechanical engineering graduate student Alex Kachkine, who restores paintings via traditional hand-painting techniques as a hobby.
He realized that many galleries have a number of paintings which never get displayed, because they require restoration that would take too long – and thus be too expensive – to perform by hand. Utilizing his method, however, restoration times could be reduced from years, months or weeks down to a matter of hours.
When developing his system, Kachkine started out with a highly damaged 15th century oil painting that he owned. Using traditional cleaning methods, he removed all the extra paint that had been applied over the years in past restoration efforts. He then performed a high-resolution scan of the cleaned painting.
Next, Alex utilized existing AI algorithms to analyze that scan, creating a digital model of what the painting likely first looked like in its pristine, unblemished form.
He subsequently used his own software to create a map of the painting which identified all the places where the original paint had faded, cracked, or was otherwise damaged. That map also indicated the exact colors that needed to be applied to those locations, in order to restore the painting to its original appearance.
Utilizing a commercial high-fidelity inkjet printer, that digital map was converted into a physical two-layer mask printed onto an ultra-thin clear polymer film. One layer was printed in the required colors – in their required locations – while the other layer was white. “In order to fully reproduce color, you need both white and color ink to get the full spectrum,” he explains.
In a final step, the mask was carefully aligned with and applied to the surface of the painting, then adhered in place with a thin spray coating of varnish.
Overview of Physically-Applied Digital Restoration
Importantly, both the mask and the varnish can be dissolved/removed with existing conservation chemicals if desired, without damaging the original paint. Additionally, the digital map will serve as a permanent record of the changes that were made to the original, for reference by future conservators.
For Alex’s project, a total of 57,314 different colors were used to repair 5,612 separate regions of the painting, in a process that took only about three and a half hours. He estimates that performing the same task solely by hand would take 66 times longer.
“There is a lot of damaged art in storage that might never be seen,” says Kachkine. “Hopefully with this new method, there’s a chance we’ll see more art, which I would be delighted by.”
A paper on the research was recently published in the journal Nature.
Source: MIT
