Art conservation requires steady hands and a keen eye. For centuries, restorers have rejuvenated artworks by pinpointing areas in need of repair, subsequently blending a precise hue to touch up one section at a time. Often, a single artwork may contain thousands of minute sections necessitating individual care. Restoring a solitary painting can take anywhere from several weeks to more than a decade.
In recent times, digital restoration technologies have paved a way for crafting virtual representations of original, restored artworks. These technologies employ methods from computer vision, image analysis, and chromatic matching to produce a “digitally restored” iteration of a painting in a relatively short span.
However, there has been no previous means of translating digital restorations directly onto an original piece until now. In a study published today in the journal Nature, Alex Kachkine, a graduate student in mechanical engineering at MIT, introduces a novel technique he has formulated to physically transfer a digital restoration directly onto an original artwork.
The restoration is printed on an exceptionally thin polymer film, shaped as a mask that can be aligned and affixed to an original piece. It can also be easily detached. Kachkine notes that a digital record of the mask can be archived and accessible for future conservators, to showcase exactly what modifications were made to restore the original artwork.
“Since there’s a digital record of which mask was utilized, in a century, when someone works on this, they’ll possess an extremely clear insight into what was executed on the painting,” Kachkine remarks. “And that’s never truly been feasible in conservation before.”
As a demonstration, he employed the approach on a seriously damaged 15th-century oil painting. The method automatically detected 5,612 distinct sections requiring restoration and filled these areas using 57,314 varied hues. The entire operation, from commencement to conclusion, took 3.5 hours, which he estimates is around 66 times quicker than conventional restoration methods.
Kachkine acknowledges that, akin to any restoration endeavor, ethical considerations must be taken into account regarding whether a restored version aptly represents an artist’s original style and intention. He states that any application of his newfound technique should be conducted in collaboration with conservators familiar with a painting’s background and origins.
“There exists a considerable amount of damaged artwork in storage that might never be showcased,” Kachkine mentions. “With this new technique, hopefully, there’s a possibility we’ll see more art, which would greatly please me.”
Digital connections
The novel restoration technique commenced as a side endeavor. In 2021, as Kachkine journeyed to MIT to begin his PhD studies in mechanical engineering, he traveled up the East Coast, deliberately visiting as many art galleries as possible along the way.
“I’ve had an affinity for art for quite a long time, since childhood,” states Kachkine, who enjoys restoring paintings as a pastime, employing traditional hand-painting methods. During his gallery visits, he recognized that the art displayed is merely a fraction of the pieces that galleries possess. Much of the art acquired by galleries is stored away due to age or damage and requires time to be properly restored.
“Restoring a painting is enjoyable, and it’s lovely to sit down and fill in gaps, concluding with a pleasant evening,” Kachkine comments. “However, it’s a remarkably slow process.”
As he has discovered, digital instruments can greatly accelerate the restoration process. Researchers have devised artificial intelligence algorithms that swiftly sift through vast amounts of data. These algorithms recognize patterns within this visual data, which they utilize to create a digitally restored representation of a specific painting, closely aligning with an artist’s style or time period. However, such digital restorations are typically showcased virtually or printed as standalone pieces and cannot be directly applied to retouch original artworks.
“All of this led me to ponder: If we could merely restore a painting digitally and enact the results physically, we would alleviate many pain points and limitations of a traditional manual process,” Kachkine reflects.
“Align and restore”
For the recent study, Kachkine devised a strategy to physically transfer a digital restoration onto an original artwork, using a 15th-century piece he acquired upon arriving at MIT. His innovative method starts with employing traditional techniques to cleanse a painting and eliminate any previous restoration attempts.
“This painting is nearly 600 years old and has undergone conservation several times,” he remarks. “In this instance, there was considerable overpainting, all of which needs to be cleaned off to discern what’s genuinely there from the outset.”
He scanned the cleaned artwork, documenting the numerous areas where paint had faded or cracked. He then utilized existing artificial intelligence algorithms to evaluate the scan and create a virtual version of how the painting likely appeared in its initial condition.
Next, Kachkine developed software that constructs a map of sections on the original artwork necessitating infilling, along with the precise colors needed to match the digitally restored depiction. This map is translated into a physical, two-layer mask printed on thin polymer-based films. The first layer is printed in color, while the second layer matches the same pattern but is printed in white.
“To fully replicate color, both white and color ink are required to achieve the full spectrum,” Kachkine clarifies. “If those two layers are misaligned, it’s incredibly obvious. Therefore, I also devised several computational tools, grounded in our understanding of human color perception, to ascertain how small a region we can feasibly align and restore.”
Kachkine utilized high-fidelity commercial inkjet printers to produce the two layers of the mask, which he meticulously aligned and layered by hand onto the original artwork, adhering them with a light spray of traditional varnish. The printed films consist of materials that can easily dissolve with conservation-grade solvents, should restorers need to expose the original, damaged art. The digital file of the mask can also be retained as a detailed account of what was restored.
For the artwork Kachkine employed, the method successfully filled thousands of losses within only a few hours. “A few years back, I was restoring a baroque Italian painting with a similar magnitude of losses, and it required me nine months of part-time effort,” he reminisces. “The more losses there are, the more advantageous this method becomes.”
He estimates that the new technique can operate orders of magnitude faster than traditional, hand-painted methods. If this technique gains widespread adoption, he stresses that conservators should play a role at every stage of the process to ensure the final artwork aligns with an artist’s style and purpose.
“This will necessitate a great deal of contemplation regarding the ethical challenges involved at every step to ascertain how this can be employed in a manner most consistent with conservation principles,” he asserts. “We’re establishing a framework for developing additional methods. As others progress with this, we will ultimately have methods that are more precise.”
This research was partially funded by the John O. and Katherine A. Lutz Memorial Fund. The study was conducted, in part, through the utilization of equipment and facilities at MIT.Nano, with supplemental support from the MIT Microsystems Technology Laboratories, the MIT Department of Mechanical Engineering, and the MIT Libraries.