Restoration of art takes permanent hands and a demanding eye. For centuries, conservators have restored images, identifying areas requiring repair, and then mixing the exact shade to fill one area at once. Often, the image can have thousands of small regions that require individual attention. Restoration of one picture may take from several weeks to over a decade.
In recent years, digital reconstruction tools have opened the way to create virtual representations of original, restored works. These tools use computer vision techniques, image recognition and color matching to generate a “digitally restored” version of the image relatively quickly.
Despite this, it was not possible to translate digital additions directly into the original work so far. IN paper appearing today in the journal NatureAlex Kachkine, a graduate of mechanical engineering in MIT, presents a new method, which he has developed to physically use a digital restaurant directly on the original image.
The restoration is printed on a very thin polymer foil, in the form of a mask that can be leveled and observed the original image. It can also be easily removed. Kachkine says that the digital mask file can be stored and dismissed by future conservators to see exactly what changes were introduced to restore the original image.
“Because there is a digital record of what the mask was used, in 100 years, next time someone works with it, there will be a very clear understanding of what was done in the picture,” says Kachkine. “And it had never been possible in protection before.”
As a demonstration, he used this method to highly damaged oil painting from the 15th century. The method automatically identified 5612 separate regions requiring repair and filled in these regions using 57 314 different colors. The whole process, from beginning to end, lasted 3.5 hours, which it estimates, is about 66 times faster than traditional reconstruction methods.
Kachkine admits that, as in any reconstruction project, ethical issues should be taken into account, in terms of whether the restored version is the appropriate representation of the artist's original style and intentions. He says that each application of his new method should take place in consultation with conservators with knowledge of the history and origin of the image.
“There is a lot of damaged art in a warehouse that you can never see,” says Kachkine. “I hope that thanks to this new method there is a chance that we will see more art that I will be delighted.”
Digital connections
The new reconstruction process began as a side project. In 2021, when Kachkine went to the MIT to start his doctoral program in the field of mechanical engineering, he went to the east coast and decided to visit as many art galleries as possible along the way.
“I have been in art for a long time since I was a child,” says Kachkine, who restores images as a hobby, using traditional manual painting techniques. He toured in galleries, he realized that art on the walls is only a fraction of the works that galleries. A significant part of the art purchased by galleries is stored because the work is aging or damaged, and it takes time to properly restore.
“Restoring the image is fun and it is great to sit, fill things and have a nice evening,” says Kachkine. “But this is a very slow process.”
As he found out, digital tools can significantly speed up the reconstruction process. Scientists have developed artificial intelligence algorithms that quickly comb huge amounts of data. Algorithms learn connections under these visual data that they use to generate a digitally restored version of a specific image, in a way that is very similar to the artist's style or period. However, such digital additions are usually displayed virtually or printed as independent work and cannot be directly used to retouch the original art.
“All this made me think: if we could simply restore the image digitally and physically influence the results, it would solve many pain points and disadvantages of a conventional manual process,” says Kachkine.
“Align and restore”
In the new study, Kachkine developed a method of physical use of a digital restaurant for the original picture, using the image of the 15th century, which he acquired when he first came to the myth. His new method is to first use traditional image cleaning techniques and remove every effort to restore past.
“This picture is almost 600 years old and has been protected many times,” he says. “In this case, there was a lot of gaps, all of which must be cleaned to see what really is at the beginning.”
A cleaned picture scanned, including many regions where the paint faded or crashed. Then he used existing artificial intelligence algorithms to analyze the scan and create a virtual version of how the image probably looked like in its original condition.
Then Kachkine has developed a software that creates a map of regions on the original image that requires filling, along with the exact colors needed to match the digitally restored version. This map is then translated into a physical, two -scale mask, which is printed into thin polymer -based films. The first layer is printed in color, and the second layer is printed exactly in the same pattern, but in white.
“To fully reproduce the color, you need both white and colorful ink to get a full spectrum,” explains Kachkine. “If these two layers are badly even, it is very easy to see.
Kachkine used commercial inks with high loyalty to print two layers of the mask, which he carefully leveled and applied manually on the original image and adjacent with a thin spray of conventional varnish. Printed films are made of materials that can be easily solved with the help of conservation class solutions, in the event that conservators had to reveal the original, damaged work. The digital mask file can also be saved as a detailed record of what has been restored.
In the case of the image used by Kachkine, the method was able to fill thousands of losses in just a few hours. “A few years ago I restored this Baroque Italian image from probably the same size of losses, and it took me nine months of part -time work,” he recalls. “The more losses, the better this method is.”
He estimates that the new method can be a ward of size faster than traditional, hand -painted approaches. If the method is widely used, he emphasizes that conservators should be involved at every stage of the process to make sure that the final work is in line with the artist's style and intention.
“At every stage of this process, it will take many deliberations on ethical challenges to see how it can be used in a way that is most consistent with the principles of protection,” he says. “We configure the framework for developing further methods. Because others work on it, we will end up with methods that are more precise.”
These works were partly supported by John O. and Katherine A. Lutz Memorial Fund. The research was carried out partly by using equipment and facilities in Mit. Nano, with additional support, myth microsystems technology laboratories, myth department of mechanical engineering and mit libraries.
